{ "base_model": "Ideogram 4", "media_type": "image", "count": 35, "total_mb": 9754.7, "skipped_count": 0, "modalities": { "t2i": "Text-to-image", "json_caption": "Structured JSON caption prompting", "bbox_layout": "Bounding-box layout control", "color_palette": "Color palette conditioning", "magic_prompt": "Magic Prompt (prompt upsampling)" }, "loras": [ { "name": "Lenovo UltraReal", "slug": "lenovo-ultrareal", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1662740, "version_id": 3025161, "version_name": "v1.0 Ideogram4", "source": "civitai", "url": "https://civitai.com/models/1662740", "repo_id": null, "file_path": null, "author": "Danrisi", "category": { "key": "photoreal", "label": "Photorealism" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": { "LoRA strength": "0.8", "CFG / guidance": "3", "Sampler": "euler" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Widely adopted — 161,776 downloads. [source: CivitAI stats]", "Positively received — 4,876 thumbs-up, 9 down. [source: CivitAI stats]", "Compact at 101 MB — cheap to stack with other LoRAs. [source: file size]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "Author labels this experimental — results vary shot to shot. [source: description]", "Author references NSFW output — review before using in shared work. [source: description]", "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "lenovo_ideogram4.safetensors", "size_bytes": 105892520, "size_mb": 100.99, "sha256": "29ab3ca98bf13eebd3ace30675878e51f3f61776e86e6f99ff601ba8dc673bd7" }, "stats": { "downloads": 161776, "thumbs_up": 4876, "thumbs_down": 9 }, "popularity": { "downloads": 161776, "thumbs_up": 4876, "thumbs_down": 9, "comments": 0, "approval": "100% positive (4876/4885)", "age_days": 85.1, "downloads_per_day": 84.06, "rank_by_downloads": 1, "rank_by_momentum": 1, "of_total": 35, "adoption_tier": "Top tier", "momentum": "Hot" }, "selection_score": 80.6, "selection_reasons": [ "161,776 downloads", "4,876 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1248x1872", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1248x1872", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1360", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1776x1184", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1776x1184", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1248x1872", "prompt": "" } ], "tags": [ "realistic", "girls", "dramatic", "photorealistic", "style" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "lenovo-ultrareal", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/lenovo-ultrareal", "local_weights_removed": true, "bytes_freed": 105892520 }, "path": "ideogram-4/lenovo-ultrareal", "archive": { "comments_total": 30, "comments": [ { "author": "2637221", "score": 4, "text": "good time to be alive, thank youu", "created_at": "2026-06-12T02:23:17.862Z" }, { "author": "810901", "score": 4, "text": "So how is the lora training with Ideogram? Did you use json captions for the dataset?", "created_at": "2026-06-12T11:55:42.400Z" }, { "author": "3122589", "score": 3, "text": "The Ideogram 4 version is fantastic! Add more examples of outputs (people preferably) to further enhance this incredible work.", "created_at": "2026-06-12T18:58:14.403Z" }, { "author": "3141503", "score": 4, "text": "Ideogram4NoiseRescale how to find it?", "created_at": "2026-06-12T19:02:44.136Z" }, { "author": "4051902", "score": 4, "text": "Woah Ideogram4 looks marvelous .", "created_at": "2026-06-13T10:06:41.380Z" }, { "author": "1091696", "score": 3, "text": "Great model man, makes anima really fun for realism with your checkpoint :)", "created_at": "2026-06-13T15:18:17.867Z" }, { "author": "12234608", "score": 7, "text": "yes i realize this is for ideogram, but it actually works great in combination with Jib Mix Realistic ZIT fp16", "created_at": "2026-06-16T00:37:45.528Z" }, { "author": "1921978", "score": 4, "text": "how do you recommend set ideogram lora strength on unconditional?", "created_at": "2026-06-17T02:01:50.073Z" }, { "author": "3537509", "score": 12, "text": "They finally added Ideogram as model option btw c:", "created_at": "2026-06-17T02:18:05.848Z" }, { "author": "8415909", "score": 0, "text": "Had a problem:\ngit cherry-pick 32ae38f1\n-> error:\nfatal: bad revision '32ae38f1'", "created_at": "2026-06-22T00:14:03.846Z" }, { "author": "4051902", "score": 0, "text": "GGUFLoaderKJ\nValueError: This model is not currently supported - (Unknown model architecture!)", "created_at": "2026-06-22T07:28:54.636Z" }, { "author": "5104492", "score": 2, "text": "Any reason strengths for conditional/unconditional not advertised?", "created_at": "2026-06-22T17:05:56.526Z" }, { "author": "67683", "score": 2, "text": "Do you plan to do one of these for Krea 2? Big fan of the model so far.", "created_at": "2026-06-24T01:09:36.715Z" }, { "author": "2823912", "score": 6, "text": "Would love to see a Krea 2 version at some point! I love your work.", "created_at": "2026-06-24T03:38:54.793Z" }, { "author": "7962800", "score": 2, "text": "yes! the king of realism loras has arrived!", "created_at": "2026-06-27T17:40:03.975Z" }, { "author": "6677484", "score": 5, "text": "this guy is a blessing", "created_at": "2026-06-27T17:55:51.203Z" } ], "faq": [ { "question": "What is Lenovo UltraReal?", "answer": "Lenovo UltraReal is a style LoRA designed for Ideogram 4.0 checkpoints, published by [Danrisi](/users/Danrisi) on CivitAI in June 2025, now on version 12. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 7,151 downloads. It's intended for photorealistic, girls, realistic, and dramatic content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "photorealistic", "style", "girls", "realistic", "dramatic" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "Danrisi", "url": "https://civarchive.com/models/1662740?modelVersionId=3025161" } }, { "name": "DnD Darkest Fantasy", "slug": "dnd-darkest-fantasy", "base_model": "Ideogram 4", "media_type": "image", "model_id": 637882, "version_id": 3107307, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/637882", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "dnddarkestfantasy" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Widely adopted — 15,038 downloads. [source: CivitAI stats]", "Positively received — 1,199 thumbs-up, 1 down. [source: CivitAI stats]", "Explicit trigger word(s) — dnddarkestfantasy — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author labels this experimental — results vary shot to shot. [source: description]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "DarkestDnD_Ideogram4.safetensors", "size_bytes": 211734008, "size_mb": 201.93, "sha256": "328ee764278d50f1009098020467163ff93455744c003386dad8cbd09e11ead7" }, "stats": { "downloads": 15038, "thumbs_up": 1199, "thumbs_down": 1 }, "popularity": { "downloads": 15038, "thumbs_up": 1199, "thumbs_down": 1, "comments": 14, "approval": "100% positive (1199/1200)", "age_days": 59.6, "downloads_per_day": 1.95, "rank_by_downloads": 2, "rank_by_momentum": 19, "of_total": 35, "adoption_tier": "Top tier", "momentum": "Long tail" }, "selection_score": 63.9, "selection_reasons": [ "15,038 downloads", "1,199 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "03_image.webp", "type": "image", "resolution": "768x1376", "prompt": "", "source": "civarchive", "civarchive_id": 136008303 } ], "tags": [ "style", "dnd", "arsmjstyle", "dark fantasy", "midjourney", "fantasy" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "dnd-darkest-fantasy", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/dnd-darkest-fantasy", "local_weights_removed": true, "bytes_freed": 211734008 }, "path": "ideogram-4/dnd-darkest-fantasy", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is DnD Darkest Fantasy?", "answer": "DnD Darkest Fantasy is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in April 2025, now on version 16. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 116 downloads. It's intended for dnd, midjourney, dark fantasy, fantasy, and arsmjstyle content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "dnddarkestfantasy" ], "tags": [ "style", "dnd", "midjourney", "dark fantasy", "fantasy", "arsmjstyle" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/637882?modelVersionId=3107307" } }, { "name": "Ghibli style (Kiki's Delivery Service)", "slug": "ghibli-style-kiki-s-delivery-service", "base_model": "Ideogram 4", "media_type": "image", "model_id": 523485, "version_id": 3045415, "version_name": "Ideogram4 v1", "source": "civitai", "url": "https://civitai.com/models/523485", "repo_id": null, "file_path": null, "author": "Yofaraway", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "Studio Ghibli style" ], "recommended_settings": { "LoRA strength": "1.3", "Sampler steps": "8", "Resolution": "768x1152", "CFG / guidance": "1", "Sampler": "Euler" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Widely adopted — 8,927 downloads. [source: CivitAI stats]", "Positively received — 604 thumbs-up, 1 down. [source: CivitAI stats]", "Explicit trigger word(s) — Studio Ghibli style — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 268 MB — cheap to stack with other LoRAs. [source: file size]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "kiki_id4.safetensors", "size_bytes": 280855400, "size_mb": 267.84, "sha256": "598b9ca7fc7d3de16750a8bb0f7bbb9ea97c5554c853d035bf0d7b24d24deb93" }, "stats": { "downloads": 8927, "thumbs_up": 604, "thumbs_down": 1 }, "popularity": { "downloads": 8927, "thumbs_up": 604, "thumbs_down": 1, "comments": 0, "approval": "100% positive (604/605)", "age_days": 78.8, "downloads_per_day": 3.55, "rank_by_downloads": 3, "rank_by_momentum": 9, "of_total": 35, "adoption_tier": "Top tier", "momentum": "Trending" }, "selection_score": 80.0, "selection_reasons": [ "8,927 downloads", "604 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" } ], "tags": [ "kiki's delivery service", "style", "ghibli", "anime" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "ghibli-style-kiki-s-delivery-service", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/ghibli-style-kiki-s-delivery-service", "local_weights_removed": true, "bytes_freed": 280855400 }, "path": "ideogram-4/ghibli-style-kiki-s-delivery-service", "archive": { "comments_total": 2, "comments": [ { "author": "13503", "score": 2, "text": "Impressed with Ideogram images you made here, it's even better than Qwen version.\nFirst strange thing : I found this Ideogram lora by pure luck, I can't find it using Search in Civitai.\nSecond strange thing : Surprisingly, the style really come out at lower quality (Turbo, 12 steps), I didn't expect that", "created_at": "2026-06-19T06:21:01.150Z" }, { "author": "2745271", "score": 2, "text": "Ideogram 4 JSON prompts?", "created_at": "2026-06-19T10:18:25.986Z" } ], "faq": [ { "question": "What is Ghibli style (Kiki's Delivery Service)?", "answer": "Ghibli style (Kiki's Delivery Service) is a style LoRA designed for Ideogram 4.0 checkpoints, published by [Yofaraway](/users/Yofaraway) on CivitAI in April 2025, now on version 11. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 280 downloads. It's intended for kiki's delivery service, ghibli, and anime content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "Studio Ghibli style" ], "tags": [ "kiki's delivery service", "style", "ghibli", "anime" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "Yofaraway", "url": "https://civarchive.com/models/523485?modelVersionId=3045415" } }, { "name": "Faceless Gods", "slug": "faceless-gods", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1951227, "version_id": 3098130, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/1951227", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Widely adopted — 6,794 downloads. [source: CivitAI stats]", "Positively received — 741 thumbs-up with no down-votes. [source: CivitAI stats]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "FacelessGodsideogram_000004000.safetensors", "size_bytes": 211734008, "size_mb": 201.93, "sha256": "c4bfb8157a7e5df1513bbbd47defc2cf807771f3623a174eb4ac0eadbb238e46" }, "stats": { "downloads": 6794, "thumbs_up": 741, "thumbs_down": 0 }, "popularity": { "downloads": 6794, "thumbs_up": 741, "thumbs_down": 0, "comments": 0, "approval": "100% positive (741/741)", "age_days": 55.7, "downloads_per_day": 1.13, "rank_by_downloads": 4, "rank_by_momentum": 23, "of_total": 35, "adoption_tier": "Top tier", "momentum": "Long tail" }, "selection_score": 62.7, "selection_reasons": [ "6,794 downloads", "741 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" } ], "tags": [ "faceless", "gods", "style" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "faceless-gods", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/faceless-gods", "local_weights_removed": true, "bytes_freed": 211734008 }, "path": "ideogram-4/faceless-gods", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Faceless Gods?", "answer": "Faceless Gods is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in September 2025, now on version 17. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 63 downloads. It's intended for faceless and gods content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "faceless", "gods", "style" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/1951227?modelVersionId=3098130" } }, { "name": "Japanese photo 1980s. Style (日本の1980年代の写真集)", "slug": "japanese-photo-1980s-style-1980", "base_model": "Ideogram 4", "media_type": "image", "model_id": 646686, "version_id": 3045521, "version_name": "v1.0 | ideogram4", "source": "civitai", "url": "https://civitai.com/models/646686", "repo_id": null, "file_path": null, "author": "allpleoleo439", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Widely adopted — 5,901 downloads. [source: CivitAI stats]", "Positively received — 429 thumbs-up with no down-votes. [source: CivitAI stats]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "idegoram480.safetensors", "size_bytes": 211733992, "size_mb": 201.93, "sha256": "cbab8aaed5d570e13976aedc3eebf7694e76e31952e28e39fd0ee94acaf83712" }, "stats": { "downloads": 5901, "thumbs_up": 429, "thumbs_down": 0 }, "popularity": { "downloads": 5901, "thumbs_up": 429, "thumbs_down": 0, "comments": 7, "approval": "100% positive (429/429)", "age_days": 78.8, "downloads_per_day": 2.12, "rank_by_downloads": 5, "rank_by_momentum": 16, "of_total": 35, "adoption_tier": "High", "momentum": "Steady" }, "selection_score": 66.7, "selection_reasons": [ "5,901 downloads", "429 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1376x768", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1376x768", "prompt": "" }, { "file": "05_image.webp", "type": "image", "resolution": "768x1376", "prompt": "", "source": "civarchive", "civarchive_id": 134121140 }, { "file": "06_image.webp", "type": "image", "resolution": "768x1376", "prompt": "", "source": "civarchive", "civarchive_id": 134121141 } ], "tags": [ "1980's", "style", "japanese" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "japanese-photo-1980s-style-1980", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/japanese-photo-1980s-style-1980", "local_weights_removed": true, "bytes_freed": 211733992 }, "path": "ideogram-4/japanese-photo-1980s-style-1980", "archive": { "comments_total": 1, "comments": [ { "author": "12234599", "score": 4, "text": "Thanks for the Ideogram version! Would love to see more of your loras for it.", "created_at": "2026-06-18T08:28:27.936Z" } ], "faq": [ { "question": "What is Japanese photo 1980s. Style (日本の1980年代の写真集)?", "answer": "Japanese photo 1980s. Style (日本の1980年代の写真集) is a style LoRA designed for Ideogram 4.0 checkpoints, published by [allpleoleo439](/users/allpleoleo439) on CivitAI in April 2025, now on version 8. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 167 downloads. It's intended for japanese and 1980's content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "japanese", "style", "1980's" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "allpleoleo439", "url": "https://civarchive.com/models/646686?modelVersionId=3045521" } }, { "name": "Dark Fantasy Pulp Pinup", "slug": "dark-fantasy-pulp-pinup", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1550211, "version_id": 3162956, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/1550211", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "Dark Fantasy Pulp Pinup," ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 4,562 downloads. [source: CivitAI stats]", "Positively received — 371 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — Dark Fantasy Pulp Pinup, — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "DarkFantasyPulpPinupIdeogram4.safetensors", "size_bytes": 211734024, "size_mb": 201.93, "sha256": "302915c5bbded753b10849c6d700191bcb5867a0a3612f4f6c3dca76b477acea" }, "stats": { "downloads": 4562, "thumbs_up": 371, "thumbs_down": 0 }, "popularity": { "downloads": 4562, "thumbs_up": 371, "thumbs_down": 0, "comments": 0, "approval": "100% positive (371/371)", "age_days": 36.5, "downloads_per_day": 2.57, "rank_by_downloads": 6, "rank_by_momentum": 13, "of_total": 35, "adoption_tier": "High", "momentum": "Steady · fresh release" }, "selection_score": 73.3, "selection_reasons": [ "4,562 downloads", "371 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "02_image.webp", "type": "image", "resolution": "1024x1024", "prompt": "", "source": "civarchive", "civarchive_id": 137740624 } ], "tags": [ "style", "olivia de berardinis", "dark fantasy", "frank frazetta", "pin-up", "luis royo" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "dark-fantasy-pulp-pinup", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/dark-fantasy-pulp-pinup", "local_weights_removed": true, "bytes_freed": 211734024 }, "path": "ideogram-4/dark-fantasy-pulp-pinup", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Dark Fantasy Pulp Pinup?", "answer": "Dark Fantasy Pulp Pinup is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in May 2025, now on version 11. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 94 downloads. It's intended for pin-up, frank frazetta, luis royo, olivia de berardinis, and dark fantasy content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "Dark Fantasy Pulp Pinup," ], "tags": [ "style", "pin-up", "frank frazetta", "luis royo", "olivia de berardinis", "dark fantasy" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/1550211?modelVersionId=3162956" } }, { "name": "Delicate Whimsy", "slug": "delicate-whimsy", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1124747, "version_id": 3075630, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/1124747", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [ "color_palette" ], "primary_modality": "color_palette", "checkpoint_family": "transformer", "trigger_words": [ "ArsMJStyle, Delicate Whimsy," ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The main 9.3B DiT and what essentially every Ideogram 4 LoRA targets. diffusers' README_ideogram4.md is explicit: \"This trainer adds LoRA to the conditional `transformer` only; the unconditional one stays frozen.\" In ComfyUI this is `ideogram4_fp8_scaled.safetensors` (9.28 GB) or its int8_convrot / nvfp4_mixed equivalents." ] }, "pros": [ "Well established — 4,019 downloads. [source: CivitAI stats]", "Positively received — 374 thumbs-up, 1 down. [source: CivitAI stats]", "Explicit trigger word(s) — ArsMJStyle, Delicate Whimsy, — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [], "file": { "name": "DelicateWhimsyideogram4.safetensors", "size_bytes": 211734016, "size_mb": 201.93, "sha256": "8ad1b44fb7eca545def96fca58dcb89f8cf1c81db403fe4d1bd6675ed6c1bc92" }, "stats": { "downloads": 4019, "thumbs_up": 374, "thumbs_down": 1 }, "popularity": { "downloads": 4019, "thumbs_up": 374, "thumbs_down": 1, "comments": 0, "approval": "100% positive (374/375)", "age_days": 69.4, "downloads_per_day": 1.83, "rank_by_downloads": 7, "rank_by_momentum": 20, "of_total": 35, "adoption_tier": "High", "momentum": "Long tail" }, "selection_score": 56.9, "selection_reasons": [ "4,019 downloads", "374 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "style", "storybook", "midjourney", "fine art", "lineart" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "delicate-whimsy", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/delicate-whimsy", "local_weights_removed": true, "bytes_freed": 211734016 }, "path": "ideogram-4/delicate-whimsy", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Delicate Whimsy?", "answer": "Delicate Whimsy is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in April 2025, now on version 11. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 127 downloads. It's intended for storybook, midjourney, fine art, and lineart content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "ArsMJStyle, Delicate Whimsy," ], "tags": [ "style", "storybook", "midjourney", "fine art", "lineart" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/1124747?modelVersionId=3075630" } }, { "name": "Tintin comic style", "slug": "tintin-comic-style", "base_model": "Ideogram 4", "media_type": "image", "model_id": 984632, "version_id": 3068405, "version_name": "Ideogram4 v1", "source": "civitai", "url": "https://civitai.com/models/984632", "repo_id": null, "file_path": null, "author": "Yofaraway", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "herge_comic" ], "recommended_settings": { "LoRA strength": "0.8", "Sampler steps": "8", "Resolution": "768x1152", "CFG / guidance": "1", "Sampler": "Euler simple" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 3,626 downloads. [source: CivitAI stats]", "Positively received — 332 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — herge_comic — so the effect can be turned on and off from the prompt. [source: model metadata]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Large at 1071 MB — slower to load and heavier to stack. [source: file size]" ], "file": { "name": "tintin_comic_style_id4.safetensors", "size_bytes": 1123105208, "size_mb": 1071.08, "sha256": "1217ff21c1250cc54e13c5fd2dd187ad43bc3ee676c90b95fc9382c046a947fe" }, "stats": { "downloads": 3626, "thumbs_up": 332, "thumbs_down": 0 }, "popularity": { "downloads": 3626, "thumbs_up": 332, "thumbs_down": 0, "comments": 0, "approval": "100% positive (332/332)", "age_days": 71.7, "downloads_per_day": 2.15, "rank_by_downloads": 8, "rank_by_momentum": 15, "of_total": 35, "adoption_tier": "High", "momentum": "Steady" }, "selection_score": 70.6, "selection_reasons": [ "3,626 downloads", "332 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "944x1408", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" } ], "tags": [ "style", "bande dessinée", "hergé", "franco belge", "comic", "bd", "tintin et milou" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "tintin-comic-style", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/tintin-comic-style", "local_weights_removed": true, "bytes_freed": 1123105208 }, "path": "ideogram-4/tintin-comic-style", "archive": { "comments_total": 1, "comments": [ { "author": "11582662", "score": 1, "text": "J'avoue ! Ce LoRA a l'air magnifique. Hâte de rentrer à la maison pour le tester ;)", "created_at": "2026-06-25T12:08:40.996Z" } ], "faq": [ { "question": "What is Tintin comic style?", "answer": "Tintin comic style is a style LoRA designed for Ideogram 4.0 checkpoints, published by [Yofaraway](/users/Yofaraway) on CivitAI in April 2025, now on version 8. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 154 downloads. It's intended for bande dessinée, hergé, franco belge, comic, and bd content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "herge_comic" ], "tags": [ "style", "bande dessinée", "hergé", "franco belge", "comic", "bd", "tintin et milou" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "Yofaraway", "url": "https://civarchive.com/models/984632?modelVersionId=3068405" } }, { "name": "70s Giallo thriller", "slug": "70s-giallo-thriller", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1743845, "version_id": 3184410, "version_name": "ideogram4", "source": "civitai", "url": "https://civitai.com/models/1743845", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 3,039 downloads. [source: CivitAI stats]", "Positively received — 234 thumbs-up with no down-votes. [source: CivitAI stats]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "70sGialloIdeogram4.safetensors", "size_bytes": 211734000, "size_mb": 201.93, "sha256": "899420437bc547f83eb335e58cef6fcefa7a6624599217ca7239c6476af2638e" }, "stats": { "downloads": 3039, "thumbs_up": 234, "thumbs_down": 0 }, "popularity": { "downloads": 3039, "thumbs_up": 234, "thumbs_down": 0, "comments": 0, "approval": "100% positive (234/234)", "age_days": 21.5, "downloads_per_day": 2.51, "rank_by_downloads": 9, "rank_by_momentum": 14, "of_total": 35, "adoption_tier": "High", "momentum": "Steady · fresh release" }, "selection_score": 71.5, "selection_reasons": [ "3,039 downloads", "234 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "giallo", "70s", "style", "thriller" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "70s-giallo-thriller", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/70s-giallo-thriller", "local_weights_removed": true, "bytes_freed": 211734000 }, "path": "ideogram-4/70s-giallo-thriller", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is 70s Giallo thriller?", "answer": "70s Giallo thriller is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in July 2025, now on version 10. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 53 downloads. It's intended for giallo, 70s, and thriller content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "giallo", "70s", "style", "thriller" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/1743845?modelVersionId=3184410" } }, { "name": "Urban Cybercore Movie", "slug": "urban-cybercore-movie", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1647065, "version_id": 3287081, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/1647065", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "Urban Cybercore Movie Still" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 2,749 downloads. [source: CivitAI stats]", "Positively received — 254 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — Urban Cybercore Movie Still — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "UrbanCybercoreIdeogram4.safetensors", "size_bytes": 211734016, "size_mb": 201.93, "sha256": "8871bd5e2216d64221fe5c03c5c073bd6218c76b166cc493760a75cabfbb2ec0" }, "stats": { "downloads": 2749, "thumbs_up": 254, "thumbs_down": 0 }, "popularity": { "downloads": 2749, "thumbs_up": 254, "thumbs_down": 0, "comments": 0, "approval": "100% positive (254/254)", "age_days": 3.5, "downloads_per_day": 9.32, "rank_by_downloads": 10, "rank_by_momentum": 3, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Hot · fresh release" }, "selection_score": 84.3, "selection_reasons": [ "2,749 downloads", "254 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "style", "cybercore" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "urban-cybercore-movie", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/urban-cybercore-movie", "local_weights_removed": true, "bytes_freed": 211734016 }, "path": "ideogram-4/urban-cybercore-movie", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Urban Cybercore Movie?", "answer": "Urban Cybercore Movie is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in June 2025, now on version 6. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 33 downloads. It's intended for cybercore content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "Urban Cybercore Movie Still" ], "tags": [ "style", "cybercore" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/1647065?modelVersionId=3287081" } }, { "name": "Pokémon retro anime style (1997)", "slug": "pok-mon-retro-anime-style-1997", "base_model": "Ideogram 4", "media_type": "image", "model_id": 863109, "version_id": 3072007, "version_name": "Ideogram4 v1", "source": "civitai", "url": "https://civitai.com/models/863109", "repo_id": null, "file_path": null, "author": "Yofaraway", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "Retro Pokémon anime style" ], "recommended_settings": { "Sampler steps": "8", "Resolution": "768x1152", "CFG / guidance": "1", "Sampler": "Euler" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 1,988 downloads. [source: CivitAI stats]", "Positively received — 156 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — Retro Pokémon anime style — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 268 MB — cheap to stack with other LoRAs. [source: file size]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "retro_pokémon_id4.safetensors", "size_bytes": 280855336, "size_mb": 267.84, "sha256": "1d1846acb82cb6c7679940b222112de5f2da4bfde8f10324fdc1e1e10d4dc117" }, "stats": { "downloads": 1988, "thumbs_up": 156, "thumbs_down": 0 }, "popularity": { "downloads": 1988, "thumbs_up": 156, "thumbs_down": 0, "comments": 0, "approval": "100% positive (156/156)", "age_days": 70.4, "downloads_per_day": 1.97, "rank_by_downloads": 11, "rank_by_momentum": 18, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Steady" }, "selection_score": 65.0, "selection_reasons": [ "1,988 downloads", "156 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1568x880", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "880x1568", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1568x880", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1408x944", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "944x1408", "prompt": "" } ], "tags": [ "vintage", "style", "pokemon", "retro", "anime", "pikachu" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "pok-mon-retro-anime-style-1997", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/pok-mon-retro-anime-style-1997", "local_weights_removed": true, "bytes_freed": 280855336 }, "path": "ideogram-4/pok-mon-retro-anime-style-1997", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Pokémon retro anime style (1997)?", "answer": "Pokémon retro anime style (1997) is a style LoRA designed for Ideogram 4.0 checkpoints, published by [Yofaraway](/users/Yofaraway) on CivitAI in April 2025, now on version 6. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 139 downloads. It's intended for vintage, pokemon, retro, anime, and pikachu content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "Retro Pokémon anime style" ], "tags": [ "vintage", "style", "pokemon", "retro", "anime", "pikachu" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "Yofaraway", "url": "https://civarchive.com/models/863109?modelVersionId=3072007" } }, { "name": "Outlined Vibrant Illustration", "slug": "outlined-vibrant-illustration", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2281260, "version_id": 3077724, "version_name": "Ideogram4 v1", "source": "civitai", "url": "https://civitai.com/models/2281260", "repo_id": null, "file_path": null, "author": "Yofaraway", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "ovi_style" ], "recommended_settings": { "LoRA strength": "1.1", "Sampler steps": "10", "Resolution": "768x1152", "CFG / guidance": "1", "Sampler": "Euler" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 1,966 downloads. [source: CivitAI stats]", "Positively received — 172 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — ovi_style — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 268 MB — cheap to stack with other LoRAs. [source: file size]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "ovi_style_id4.safetensors", "size_bytes": 280855384, "size_mb": 267.84, "sha256": "343ebbdfb39b3f9f30c72f1045724ccb798db885a8c362b47088156aeb068e79" }, "stats": { "downloads": 1966, "thumbs_up": 172, "thumbs_down": 0 }, "popularity": { "downloads": 1966, "thumbs_up": 172, "thumbs_down": 0, "comments": 0, "approval": "100% positive (172/172)", "age_days": 66.6, "downloads_per_day": 2.01, "rank_by_downloads": 12, "rank_by_momentum": 17, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Steady" }, "selection_score": 65.5, "selection_reasons": [ "1,966 downloads", "172 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1536", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "4096x2736", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "944x1408", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "768x1152", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "2048x1152", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "944x1408", "prompt": "" } ], "tags": [ "voyager spirit", "style" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "outlined-vibrant-illustration", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/outlined-vibrant-illustration", "local_weights_removed": true, "bytes_freed": 280855384 }, "path": "ideogram-4/outlined-vibrant-illustration", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Outlined Vibrant Illustration?", "answer": "Outlined Vibrant Illustration is a style LoRA designed for Ideogram 4.0 checkpoints, published by [Yofaraway](/users/Yofaraway) on CivitAI in January 2026, now on version 6. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 134 downloads. It's intended for voyager spirit content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "ovi_style" ], "tags": [ "voyager spirit", "style" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "Yofaraway", "url": "https://civarchive.com/models/2281260?modelVersionId=3077724" } }, { "name": "Ideogram 4 TurboTime", "slug": "ideogram-4-turbotime", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2711950, "version_id": 3046680, "version_name": "v1.0", "source": "civitai", "url": "https://civitai.com/models/2711950", "repo_id": null, "file_path": null, "author": "ostris", "category": { "key": "accelerator", "label": "Speed / step-reduction" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": { "Sampler steps": "2" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "One real step-reduction LoRA exists: **TurboTime** by ostris (`ostris/ideogram_4_turbotime_lora`, file `ideogram_4_turbotime_v1.safetensors`, 846,778,576 bytes ≈ 808 MiB; also mirrored on CivitAI under the same filename). The model card: it \"is capable of generating good quality images in as little as 2 steps with no CFG and no unconditional model\", made with an in-progress \"continuious turbo training method\", with updated versions likely. The card's own sample grid is at 8 steps. The nomadoor community workflow describes it as a \"2 to 8 step LoRA\" and runs it at 8 steps, cfg 1.0, strength 0.6", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 1,679 downloads. [source: CivitAI stats]", "Positively received — 98 thumbs-up with no down-votes. [source: CivitAI stats]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Large at 808 MB — slower to load and heavier to stack. [source: file size]" ], "file": { "name": "ideogram_4_turbotime_v1.safetensors", "size_bytes": 846778576, "size_mb": 807.55, "sha256": "e44a95e54b417d90492538f0bdfdb73be4f2234399035a10fa53f534399672ef" }, "stats": { "downloads": 1679, "thumbs_up": 98, "thumbs_down": 0 }, "popularity": { "downloads": 1679, "thumbs_up": 98, "thumbs_down": 0, "comments": 15, "approval": "100% positive (98/98)", "age_days": 78.5, "downloads_per_day": 21.39, "rank_by_downloads": 13, "rank_by_momentum": 2, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Hot" }, "selection_score": 74.1, "selection_reasons": [ "1,679 downloads", "98 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1248x1872", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1248x1872", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1824x1248", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "768x1280", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1248x1872", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1712", "prompt": "" } ], "tags": [ "turbo", "tool" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "ideogram-4-turbotime", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/ideogram-4-turbotime", "local_weights_removed": true, "bytes_freed": 846778576 }, "path": "ideogram-4/ideogram-4-turbotime", "archive": { "comments_total": 18, "comments": [ { "author": "14648", "score": 2, "text": "Instantly recognised that horse dj from the training guide(s). Welcome back to civ, ostris, been a little while.", "created_at": "2026-06-18T15:59:47.237Z" }, { "author": "354458", "score": 1, "text": "Nice. Was waiting for this.\nI managed to bake in trained LoRa to both base models, instead of using two Load Lora nodes which requires heavy background calculations in ComfyUI on the fly. Such pre-merged models reduces time and required resources without any loss. With proper weights, it also eliminates the \"safety filter\". Baking in Turbo LoRa could be another great addition to both one model and two model modes.", "created_at": "2026-06-18T16:38:10.418Z" }, { "author": "1013492", "score": 7, "text": "This LoRA doesn't work with other LoRAs though. So does that mean other LoRAs will have to be trained with this LoRA in mind?", "created_at": "2026-06-18T18:19:05.286Z" }, { "author": "583639", "score": 38, "text": "it was created by ostrich, at least give him credit.", "created_at": "2026-06-18T19:06:51.883Z" }, { "author": "1493937", "score": 2, "text": "buy GPU or use lora..what do you think who win ? ..", "created_at": "2026-06-19T03:13:27.249Z" }, { "author": "1377768", "score": 5, "text": "Thanks so much for this Ostris!, Ideogram 4.0 is great, but the 150+ second generation speed was killing me, now I can make a good image in 15 seconds (on my 3090 at 3-4 steps) and maybe seed hunt and upscale to best ones.", "created_at": "2026-06-19T10:05:05.897Z" }, { "author": "8430645", "score": 0, "text": "Anyone got this working on the gguf version ? i already updated my sd-cli.exe to newest, tried 8 steps also, reults look horrible like 8 steps without lora (no difference).\nMaybe only for diffuser model ?", "created_at": "2026-06-20T07:16:21.272Z" }, { "author": "8589506", "score": 0, "text": "What strength should the model be, and do i need to have a lora for both for the uncond and postive ideogram checkpoints?", "created_at": "2026-06-21T03:09:35.107Z" }, { "author": "5857720", "score": 0, "text": "Good (necessary) initiative, but not ideal results yet. All images get very dark using it when there are other loras along.", "created_at": "2026-06-21T04:44:15.645Z" }, { "author": "12012019", "score": 0, "text": "This is great für Speed. I really like it and it it perfect for SFW Stuff. On the other hand when it comes to NSFW stuff, it seems to have conflicts and problems showing expplicit details.", "created_at": "2026-06-21T14:23:51.237Z" }, { "author": "1692275", "score": 0, "text": "Is there a workflow for this? I can't see it here or on huggingface.", "created_at": "2026-06-22T13:33:19.537Z" } ], "faq": [ { "question": "What is Ideogram 4 TurboTime?", "answer": "Ideogram 4 TurboTime is a LoRA designed for Ideogram 4.0 checkpoints, published by [ostris](/users/ostris) on CivitAI in June 2026. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 1,679 downloads. It's intended for turbo and tool content." } ], "trigger_words": [], "tags": [ "turbo", "tool" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ostris", "url": "https://civarchive.com/models/2711950?modelVersionId=3046680" } }, { "name": "zjourney - Fantasy Realism Refiner Ideogram 4/Krea 2", "slug": "zjourney-fantasy-realism-refiner-ideogram-4-krea-2", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2686999, "version_id": 3025699, "version_name": "zjourneyideogramV2", "source": "civitai", "url": "https://civitai.com/models/2686999", "repo_id": null, "file_path": null, "author": "tsolful", "category": { "key": "photoreal", "label": "Photorealism" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": { "LoRA strength": "0.8–1.4" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 1,565 downloads. [source: CivitAI stats]", "Positively received — 96 thumbs-up with no down-votes. [source: CivitAI stats]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "zjourneyv2.safetensors", "size_bytes": 423415536, "size_mb": 403.8, "sha256": "57a703e64a3b9581b79a74a61ba3e0c597bfc5906a9f13c21041c10f237984f5" }, "stats": { "downloads": 1565, "thumbs_up": 96, "thumbs_down": 0 }, "popularity": { "downloads": 1565, "thumbs_up": 96, "thumbs_down": 0, "comments": 0, "approval": "100% positive (96/96)", "age_days": 84.8, "downloads_per_day": 5.95, "rank_by_downloads": 14, "rank_by_momentum": 5, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Trending" }, "selection_score": 69.6, "selection_reasons": [ "1,565 downloads", "96 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "2160x3048", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "2464x3264", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "2688x1792", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "3938x1700", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1456x1456", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1680x1264", "prompt": "" } ], "tags": [ "fantasy", "style", "ideogram", "ideogram 4", "realism" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "zjourney-fantasy-realism-refiner-ideogram-4-krea-2", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/zjourney-fantasy-realism-refiner-ideogram-4-krea-2", "local_weights_removed": true, "bytes_freed": 423415536 }, "path": "ideogram-4/zjourney-fantasy-realism-refiner-ideogram-4-krea-2", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is zjourney - Fantasy Realism Refiner Ideogram 4/Krea 2?", "answer": "zjourney - Fantasy Realism Refiner Ideogram 4/Krea 2 is a style LoRA designed for Ideogram 4.0 checkpoints, published by [tsolful](/users/tsolful) on CivitAI in June 2026, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 505 downloads. It's intended for fantasy, ideogram, ideogram 4, and realism content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "fantasy", "style", "ideogram", "ideogram 4", "realism" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "tsolful", "url": "https://civarchive.com/models/2686999?modelVersionId=3025699" } }, { "name": "Ferrari F40 - 1987", "slug": "ferrari-f40-1987", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1825103, "version_id": 3026359, "version_name": "v1.0 IDEOGRAM4", "source": "civitai", "url": "https://civitai.com/models/1825103", "repo_id": null, "file_path": null, "author": "mastermerlin", "category": { "key": "accelerator", "label": "Speed / step-reduction" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "FerrariF40" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "One real step-reduction LoRA exists: **TurboTime** by ostris (`ostris/ideogram_4_turbotime_lora`, file `ideogram_4_turbotime_v1.safetensors`, 846,778,576 bytes ≈ 808 MiB; also mirrored on CivitAI under the same filename). The model card: it \"is capable of generating good quality images in as little as 2 steps with no CFG and no unconditional model\", made with an in-progress \"continuious turbo training method\", with updated versions likely. The card's own sample grid is at 8 steps. The nomadoor community workflow describes it as a \"2 to 8 step LoRA\" and runs it at 8 steps, cfg 1.0, strength 0.6", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 1,139 downloads. [source: CivitAI stats]", "Positively received — 75 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — FerrariF40 — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 163 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "mastermerlin_ideogram4-fp8_Ferrari_F40_1987_final.safetensors", "size_bytes": 170503400, "size_mb": 162.6, "sha256": "1f91930d408661287aff90c819b4d7b70ce133d9e36142c008827e0041669edf" }, "stats": { "downloads": 1139, "thumbs_up": 75, "thumbs_down": 0 }, "popularity": { "downloads": 1139, "thumbs_up": 75, "thumbs_down": 0, "comments": 6, "approval": "100% positive (75/75)", "age_days": 84.6, "downloads_per_day": 0.43, "rank_by_downloads": 15, "rank_by_momentum": 33, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Long tail" }, "selection_score": 39.9, "selection_reasons": [ "1,139 downloads", "75 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1680x944", "prompt": "" } ], "tags": [ "ferrari", "vehicle", "f40" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "ferrari-f40-1987", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/ferrari-f40-1987", "local_weights_removed": true, "bytes_freed": 170503400 }, "path": "ideogram-4/ferrari-f40-1987", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Ferrari F40 - 1987?", "answer": "Ferrari F40 - 1987 is a LoRA designed for Ideogram 4.0 checkpoints, published by [mastermerlin](/users/mastermerlin) on CivitAI in August 2025, now on version 13. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 36 downloads. It's intended for vehicle, ferrari, and f40 content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "FerrariF40" ], "tags": [ "vehicle", "ferrari", "f40" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "mastermerlin", "url": "https://civarchive.com/models/1825103?modelVersionId=3026359" } }, { "name": "Porsche 911 Turbo Targa (930) - 1989", "slug": "porsche-911-turbo-targa-930-1989", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1815501, "version_id": 3031109, "version_name": "v1.0 IDEOGRAM4", "source": "civitai", "url": "https://civitai.com/models/1815501", "repo_id": null, "file_path": null, "author": "mastermerlin", "category": { "key": "accelerator", "label": "Speed / step-reduction" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "911-930T4rga" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "One real step-reduction LoRA exists: **TurboTime** by ostris (`ostris/ideogram_4_turbotime_lora`, file `ideogram_4_turbotime_v1.safetensors`, 846,778,576 bytes ≈ 808 MiB; also mirrored on CivitAI under the same filename). The model card: it \"is capable of generating good quality images in as little as 2 steps with no CFG and no unconditional model\", made with an in-progress \"continuious turbo training method\", with updated versions likely. The card's own sample grid is at 8 steps. The nomadoor community workflow describes it as a \"2 to 8 step LoRA\" and runs it at 8 steps, cfg 1.0, strength 0.6", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Well established — 1,012 downloads. [source: CivitAI stats]", "Positively received — 61 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — 911-930T4rga — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 163 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]" ], "file": { "name": "mastermerlin_ideogram4-fp8_Porsche_911_930_Turbo_33_Targa_1989_final.safetensors", "size_bytes": 170503416, "size_mb": 162.6, "sha256": "1ffa06db3aefce27c291261de86ebc1015088e49ef1f8141d49151ade5089c72" }, "stats": { "downloads": 1012, "thumbs_up": 61, "thumbs_down": 0 }, "popularity": { "downloads": 1012, "thumbs_up": 61, "thumbs_down": 0, "comments": 0, "approval": "100% positive (61/61)", "age_days": 83.1, "downloads_per_day": 0.35, "rank_by_downloads": 16, "rank_by_momentum": 35, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Long tail" }, "selection_score": 37.1, "selection_reasons": [ "1,012 downloads", "61 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1936x1088", "prompt": "" } ], "tags": [ "targa", "vehicle", "930", "turbo", "911", "porsche" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "porsche-911-turbo-targa-930-1989", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/porsche-911-turbo-targa-930-1989", "local_weights_removed": true, "bytes_freed": 170503416 }, "path": "ideogram-4/porsche-911-turbo-targa-930-1989", "archive": { "comments_total": 4, "comments": [ { "author": "1377768", "score": 1, "text": "I am starting to think this guy might like Porsches a little.", "created_at": "2026-06-14T21:33:36.952Z" }, { "author": "10791649", "score": 0, "text": "May I ask you, how you trained Loras on Lens? Can you maybe share settings and stuff….??", "created_at": "2026-06-15T10:38:28.997Z" } ], "faq": [ { "question": "What is Porsche 911 Turbo Targa (930) - 1989?", "answer": "Porsche 911 Turbo Targa (930) - 1989 is a LoRA designed for Ideogram 4.0 checkpoints, published by [mastermerlin](/users/mastermerlin) on CivitAI in August 2025, now on version 13. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 29 downloads. It's intended for porsche, vehicle, turbo, 911, and 930 content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "911-930T4rga" ], "tags": [ "porsche", "vehicle", "turbo", "911", "930", "targa" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "mastermerlin", "url": "https://civarchive.com/models/1815501?modelVersionId=3031109" } }, { "name": "Lamborghini Terzo Millenio - 2017", "slug": "lamborghini-terzo-millenio-2017", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1343068, "version_id": 3058688, "version_name": "v2.3 IDEOGRAM4", "source": "civitai", "url": "https://civitai.com/models/1343068", "repo_id": null, "file_path": null, "author": "mastermerlin", "category": { "key": "accelerator", "label": "Speed / step-reduction" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "TerzoMill3nnio" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "One real step-reduction LoRA exists: **TurboTime** by ostris (`ostris/ideogram_4_turbotime_lora`, file `ideogram_4_turbotime_v1.safetensors`, 846,778,576 bytes ≈ 808 MiB; also mirrored on CivitAI under the same filename). The model card: it \"is capable of generating good quality images in as little as 2 steps with no CFG and no unconditional model\", made with an in-progress \"continuious turbo training method\", with updated versions likely. The card's own sample grid is at 8 steps. The nomadoor community workflow describes it as a \"2 to 8 step LoRA\" and runs it at 8 steps, cfg 1.0, strength 0.6", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Explicit trigger word(s) — TerzoMill3nnio — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 163 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (991 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "mastermerlin_ideogram4-fp8_Lamborghini_Terzo_Millennio_final.safetensors", "size_bytes": 170503408, "size_mb": 162.6, "sha256": "9b7a41a7546b517f8dfc9e624e0731383d027f068e09d0d9b5db50120f347dee" }, "stats": { "downloads": 991, "thumbs_up": 48, "thumbs_down": 0 }, "popularity": { "downloads": 991, "thumbs_up": 48, "thumbs_down": 0, "comments": 0, "approval": "100% positive (48/48)", "age_days": 74.8, "downloads_per_day": 0.44, "rank_by_downloads": 17, "rank_by_momentum": 32, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Long tail" }, "selection_score": 40.1, "selection_reasons": [ "991 downloads", "48 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" } ], "tags": [ "terzo millenio", "vehicle", "lamborghini" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "lamborghini-terzo-millenio-2017", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/lamborghini-terzo-millenio-2017", "local_weights_removed": true, "bytes_freed": 170503408 }, "path": "ideogram-4/lamborghini-terzo-millenio-2017", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Lamborghini Terzo Millenio - 2017?", "answer": "Lamborghini Terzo Millenio - 2017 is a LoRA designed for Ideogram 4.0 checkpoints, published by [mastermerlin](/users/mastermerlin) on CivitAI in April 2025, now on version 13. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 33 downloads. It's intended for vehicle, lamborghini, and terzo millenio content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "TerzoMill3nnio" ], "tags": [ "vehicle", "lamborghini", "terzo millenio" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "mastermerlin", "url": "https://civarchive.com/models/1343068?modelVersionId=3058688" } }, { "name": "70s Fairytale", "slug": "70s-fairytale", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2716782, "version_id": 3065096, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/2716782", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "70sFairytale," ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Positively received — 99 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — 70sFairytale, — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (893 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "70sFairytaleIdeogram4.safetensors", "size_bytes": 211734008, "size_mb": 201.93, "sha256": "8c72823211e2cd18431aa5b24582c60f65ce943e4df773584b23215b0327feb1" }, "stats": { "downloads": 893, "thumbs_up": 99, "thumbs_down": 0 }, "popularity": { "downloads": 893, "thumbs_up": 99, "thumbs_down": 0, "comments": 0, "approval": "100% positive (99/99)", "age_days": 66.8, "downloads_per_day": 1.68, "rank_by_downloads": 18, "rank_by_momentum": 22, "of_total": 35, "adoption_tier": "Moderate", "momentum": "Long tail" }, "selection_score": 50.2, "selection_reasons": [ "893 downloads", "99 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1376x768", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1376x768", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" } ], "tags": [ "brian froud", "style", "arthur rackham" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "70s-fairytale", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/70s-fairytale", "local_weights_removed": true, "bytes_freed": 211734008 }, "path": "ideogram-4/70s-fairytale", "archive": { "comments_total": 2, "comments": [ { "author": "231866", "score": 0, "text": "70s fairy tales? Damn those were creepy AF", "created_at": "2026-06-30T13:19:38.046Z" } ], "faq": [ { "question": "What is 70s Fairytale?", "answer": "70s Fairytale is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in June 2026, now on version 7. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 112 downloads. It's intended for brian froud and arthur rackham content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "70sFairytale," ], "tags": [ "brian froud", "style", "arthur rackham" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/2716782?modelVersionId=3065096" } }, { "name": "Vintage Anime - Ideogram 4", "slug": "vintage-anime-ideogram-4", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2685958, "version_id": 3029880, "version_name": "v3.0", "source": "civitai", "url": "https://civitai.com/models/2685958", "repo_id": null, "file_path": null, "author": "kingroka", "category": { "key": "style", "label": "Visual style" }, "modalities": [ "color_palette" ], "primary_modality": "color_palette", "checkpoint_family": "transformer", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The main 9.3B DiT and what essentially every Ideogram 4 LoRA targets. diffusers' README_ideogram4.md is explicit: \"This trainer adds LoRA to the conditional `transformer` only; the unconditional one stays frozen.\" In ComfyUI this is `ideogram4_fp8_scaled.safetensors` (9.28 GB) or its int8_convrot / nvfp4_mixed equivalents." ] }, "pros": [ "Positively received — 53 thumbs-up with no down-votes. [source: CivitAI stats]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Modest adoption (716 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "vintage-anime-ideogram_v3.safetensors", "size_bytes": 211734008, "size_mb": 201.93, "sha256": "6d39652f38bdbb1e08dcc6fa26040c22fd72513aa28be0b402382cc15ca05170" }, "stats": { "downloads": 716, "thumbs_up": 53, "thumbs_down": 0 }, "popularity": { "downloads": 716, "thumbs_up": 53, "thumbs_down": 0, "comments": 2, "approval": "100% positive (53/53)", "age_days": 83.5, "downloads_per_day": 5.32, "rank_by_downloads": 19, "rank_by_momentum": 6, "of_total": 35, "adoption_tier": "Modest", "momentum": "Trending" }, "selection_score": 64.2, "selection_reasons": [ "716 downloads", "53 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "832x1216", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "2112x1072", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "832x1216", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "2400x1072", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "2112x1072", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "2112x1072", "prompt": "" } ], "tags": [ "anime", "style" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "vintage-anime-ideogram-4", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/vintage-anime-ideogram-4", "local_weights_removed": true, "bytes_freed": 211734008 }, "path": "ideogram-4/vintage-anime-ideogram-4", "archive": { "comments_total": 1, "comments": [ { "author": "1192055", "score": 3, "text": "By far the best version. Great job! It's been cool following this development.", "created_at": "2026-06-13T15:26:34.223Z" } ], "faq": [ { "question": "What is Vintage Anime - Ideogram 4?", "answer": "Vintage Anime - Ideogram 4 is a style LoRA designed for Ideogram 4.0 checkpoints, published by [kingroka](/users/kingroka) on CivitAI in June 2026, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 444 downloads. It's intended for anime content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "anime", "style" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "kingroka", "url": "https://civarchive.com/models/2685958?modelVersionId=3029880" } }, { "name": "LP++ | PSX/PS1 Inspired", "slug": "lp-psx-ps1-inspired", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2569527, "version_id": 3041629, "version_name": "V3 - Ideogram", "source": "civitai", "url": "https://civitai.com/models/2569527", "repo_id": null, "file_path": null, "author": "flowask", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": { "LoRA strength": "0.8–1.2", "Sampler steps": "8", "Resolution": "1024x102", "Sampler": "LCM" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Positively received — 57 thumbs-up with no down-votes. [source: CivitAI stats]", "Compact at 303 MB — cheap to stack with other LoRAs. [source: file size]", "Author published concrete recommended settings. [source: description]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (694 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "ERPSX.safetensors", "size_bytes": 317574840, "size_mb": 302.86, "sha256": "2fab460288c1ef921de6c82bd9a89b54118e1ef0be199876f57ed8d2282693e0" }, "stats": { "downloads": 694, "thumbs_up": 57, "thumbs_down": 0 }, "popularity": { "downloads": 694, "thumbs_up": 57, "thumbs_down": 0, "comments": 0, "approval": "100% positive (57/57)", "age_days": 80.1, "downloads_per_day": 1.79, "rank_by_downloads": 20, "rank_by_momentum": 21, "of_total": 35, "adoption_tier": "Modest", "momentum": "Long tail" }, "selection_score": 48.2, "selection_reasons": [ "694 downloads", "57 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "896x1088", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "896x1088", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "896x1088", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "ps1/psx", "low poly", "style", "aesthetic" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "lp-psx-ps1-inspired", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/lp-psx-ps1-inspired", "local_weights_removed": true, "bytes_freed": 317574840 }, "path": "ideogram-4/lp-psx-ps1-inspired", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is LP++ | PSX/PS1 Inspired?", "answer": "LP++ | PSX/PS1 Inspired is a style LoRA designed for Ideogram 4.0 checkpoints, published by [flowask](/users/flowask) on CivitAI in April 2026, now on version 5. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 143 downloads. It's intended for aesthetic, low poly, and ps1/psx content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "aesthetic", "style", "low poly", "ps1/psx" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "flowask", "url": "https://civarchive.com/models/2569527?modelVersionId=3041629" } }, { "name": "Anime / Digital Painting - Ideogram 4 LoRA", "slug": "anime-digital-painting-ideogram-4-lora", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2707294, "version_id": 3040843, "version_name": "v1.0", "source": "civitai", "url": "https://civitai.com/models/2707294", "repo_id": null, "file_path": null, "author": "kingroka", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "anime" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Positively received — 58 thumbs-up with no down-votes. [source: CivitAI stats]", "Explicit trigger word(s) — anime — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Thin documentation — expect trial and error. [source: description length]", "Modest adoption (655 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "anime-ideo4_v1.safetensors", "size_bytes": 211734000, "size_mb": 201.93, "sha256": "011854394b946ef790dd6947289a06c88d327acaf8bf0bd52f1988a50b50a5d3" }, "stats": { "downloads": 655, "thumbs_up": 58, "thumbs_down": 0 }, "popularity": { "downloads": 655, "thumbs_up": 58, "thumbs_down": 0, "comments": 4, "approval": "100% positive (58/58)", "age_days": 80.3, "downloads_per_day": 8.16, "rank_by_downloads": 21, "rank_by_momentum": 4, "of_total": 35, "adoption_tier": "Modest", "momentum": "Hot" }, "selection_score": 64.8, "selection_reasons": [ "655 downloads", "58 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "688x1024", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "832x1024", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "704x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "784x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "688x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "2112x1072", "prompt": "" } ], "tags": [ "style", "digital", "digital painting", "anime" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "anime-digital-painting-ideogram-4-lora", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/anime-digital-painting-ideogram-4-lora", "local_weights_removed": true, "bytes_freed": 211734000 }, "path": "ideogram-4/anime-digital-painting-ideogram-4-lora", "archive": { "comments_total": 6, "comments": [ { "author": "59524", "score": 0, "text": "nice, but where are the examples of prompts?", "created_at": "2026-06-16T21:41:04.550Z" }, { "author": "70332", "score": 0, "text": "Thank you! Any chance we can get a workflow?", "created_at": "2026-06-18T13:38:13.729Z" }, { "author": "10526218", "score": 0, "text": "How many steps do we need ? i can tell that 25 is not enough to generate a picture, and what scheduler is needed? I don't have the same result as you unfortunately.", "created_at": "2026-06-20T20:51:32.049Z" } ], "faq": [ { "question": "What is Anime / Digital Painting - Ideogram 4 LoRA?", "answer": "Anime / Digital Painting - Ideogram 4 LoRA is a style LoRA designed for Ideogram 4.0 checkpoints, published by [kingroka](/users/kingroka) on CivitAI in June 2026. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 655 downloads. It's intended for digital, digital painting, and anime content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "anime" ], "tags": [ "style", "digital", "digital painting", "anime" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "kingroka", "url": "https://civarchive.com/models/2707294?modelVersionId=3040843" } }, { "name": "Furry Enhancer (Ideogram4)", "slug": "furry-enhancer-ideogram4", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2695892, "version_id": 3139067, "version_name": "v2.8 (Realistic)", "source": "civitai", "url": "https://civitai.com/models/2695892", "repo_id": null, "file_path": null, "author": "freek22", "category": { "key": "detail", "label": "Detail / anatomy fixer" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": { "LoRA strength": "0.8" }, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Author published concrete recommended settings. [source: description]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (555 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "furry enhancerideogramV2.8realistic.safetensors", "size_bytes": 423415064, "size_mb": 403.8, "sha256": "8eff27870655f22d6e92426bbec4743435c8d0920a11e2c1f5e70f7fee76abcb" }, "stats": { "downloads": 555, "thumbs_up": 41, "thumbs_down": 0 }, "popularity": { "downloads": 555, "thumbs_up": 41, "thumbs_down": 0, "comments": 0, "approval": "100% positive (41/41)", "age_days": 49.4, "downloads_per_day": 3.38, "rank_by_downloads": 22, "rank_by_momentum": 10, "of_total": 35, "adoption_tier": "Modest", "momentum": "Steady" }, "selection_score": 60.9, "selection_reasons": [ "555 downloads", "41 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1248x1600", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1248x1600", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "2112x1184", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1632x1216", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1392x1792", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1248x1600", "prompt": "" } ], "tags": [ "animals", "creatures", "art", "realism", "cat", "anthro", "furry", "fantasy", "photorealism", "lion", "anthropomorphic", "leopard", "ideogram", "ideogram4", "animal" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "furry-enhancer-ideogram4", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/furry-enhancer-ideogram4", "local_weights_removed": true, "bytes_freed": 423415064 }, "path": "ideogram-4/furry-enhancer-ideogram4", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Furry Enhancer (Ideogram4)?", "answer": "Furry Enhancer (Ideogram4) is a LoRA designed for Ideogram 4.0 checkpoints, published by [freek22](/users/freek22) on CivitAI in June 2026, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 167 downloads. It's intended for animals, creatures, art, realism, and cat content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "animals", "creatures", "art", "realism", "cat", "anthro", "furry", "fantasy", "photorealism", "animal", "lion", "anthropomorphic", "leopard", "ideogram", "ideogram4" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "freek22", "url": "https://civarchive.com/models/2695892?modelVersionId=3139067" } }, { "name": "Porsche Targa [Qwen, WAN, Hidream, Ideogram]", "slug": "porsche-targa-qwen-wan-hidream-ideogram", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1490273, "version_id": 3080803, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/1490273", "repo_id": null, "file_path": null, "author": "denrakeiw", "category": { "key": "detail", "label": "Detail / anatomy fixer" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "targa" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Explicit trigger word(s) — targa — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (438 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "Targa_Ideogram_000002750.safetensors", "size_bytes": 211734048, "size_mb": 201.93, "sha256": "ac3e1f4b52466aceadffc7a71852dd652a6ded39f2e89730620d3ba13d0ef17b" }, "stats": { "downloads": 438, "thumbs_up": 42, "thumbs_down": 0 }, "popularity": { "downloads": 438, "thumbs_up": 42, "thumbs_down": 0, "comments": 5, "approval": "100% positive (42/42)", "age_days": 67.9, "downloads_per_day": 0.47, "rank_by_downloads": 23, "rank_by_momentum": 31, "of_total": 35, "adoption_tier": "Modest", "momentum": "Long tail" }, "selection_score": 36.5, "selection_reasons": [ "438 downloads", "42 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "928x1136", "prompt": "" }, { "file": "05_image.webp", "type": "image", "resolution": "1088x1936", "prompt": "", "source": "civarchive", "civarchive_id": 135211358 } ], "tags": [ "hidream", "vehicle", "porsche", "sportscar" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "porsche-targa-qwen-wan-hidream-ideogram", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/porsche-targa-qwen-wan-hidream-ideogram", "local_weights_removed": true, "bytes_freed": 211734048 }, "path": "ideogram-4/porsche-targa-qwen-wan-hidream-ideogram", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Porsche Targa [Qwen, WAN, Hidream, Ideogram]?", "answer": "Porsche Targa [Qwen, WAN, Hidream, Ideogram] is a LoRA designed for Ideogram 4.0 checkpoints, published by [denrakeiw](/users/denrakeiw) on CivitAI in April 2025, now on version 4. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 32 downloads. It's intended for porsche, vehicle, sportscar, and hidream content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "targa" ], "tags": [ "porsche", "vehicle", "sportscar", "hidream" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "denrakeiw", "url": "https://civarchive.com/models/1490273?modelVersionId=3080803" } }, { "name": "Studio Ghibli Style", "slug": "studio-ghibli-style", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2718056, "version_id": 3054263, "version_name": "v1.0", "source": "civitai", "url": "https://civitai.com/models/2718056", "repo_id": null, "file_path": null, "author": "kingroka", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "studi ghibli" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Explicit trigger word(s) — studi ghibli — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (364 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "ghibli-ideogram_v1.safetensors", "size_bytes": 211734000, "size_mb": 201.93, "sha256": "18a77c521b3824f64cb74830b4afdc974b684513888c2923d8c563248a688276" }, "stats": { "downloads": 364, "thumbs_up": 38, "thumbs_down": 0 }, "popularity": { "downloads": 364, "thumbs_up": 38, "thumbs_down": 0, "comments": 0, "approval": "100% positive (38/38)", "age_days": 76.1, "downloads_per_day": 4.78, "rank_by_downloads": 24, "rank_by_momentum": 7, "of_total": 35, "adoption_tier": "Modest", "momentum": "Trending" }, "selection_score": 59.5, "selection_reasons": [ "364 downloads", "38 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "704x1024", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "2112x1072", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "688x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1344x1152", "prompt": "" } ], "tags": [ "ghibli", "style", "secret life of arrietty", "princess mononoke", "ponyo", "spirited away", "studio ghibli" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "studio-ghibli-style", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/studio-ghibli-style", "local_weights_removed": true, "bytes_freed": 211734000 }, "path": "ideogram-4/studio-ghibli-style", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Studio Ghibli Style?", "answer": "Studio Ghibli Style is a style LoRA designed for Ideogram 4.0 checkpoints, published by [kingroka](/users/kingroka) on CivitAI in June 2026. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 364 downloads. It's intended for ghibli, secret life of arrietty, princess mononoke, ponyo, and spirited away content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "studi ghibli" ], "tags": [ "ghibli", "style", "secret life of arrietty", "princess mononoke", "ponyo", "spirited away", "studio ghibli" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "kingroka", "url": "https://civarchive.com/models/2718056?modelVersionId=3054263" } }, { "name": "MiHoYo Style", "slug": "mihoyo-style", "base_model": "Ideogram 4", "media_type": "image", "model_id": 995865, "version_id": 3061630, "version_name": "v1.0 | ideogram4", "source": "civitai", "url": "https://civitai.com/models/995865", "repo_id": null, "file_path": null, "author": "allpleoleo439", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author labels this experimental — results vary shot to shot. [source: description]", "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (309 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "idegoram4_miha_re3_000003000.safetensors", "size_bytes": 211734008, "size_mb": 201.93, "sha256": "8d6b75860935f6af02b2f89d5ae6a38e46d1312b60d12b3433db03da5c93ed86" }, "stats": { "downloads": 309, "thumbs_up": 26, "thumbs_down": 0 }, "popularity": { "downloads": 309, "thumbs_up": 26, "thumbs_down": 0, "comments": 0, "approval": "100% positive (26/26)", "age_days": 73.9, "downloads_per_day": 0.74, "rank_by_downloads": 25, "rank_by_momentum": 26, "of_total": 35, "adoption_tier": "Modest", "momentum": "Long tail" }, "selection_score": 39.3, "selection_reasons": [ "309 downloads", "26 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1680x1264", "prompt": "" } ], "tags": [ "china", "style", "anime", "game", "mmorpg" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "mihoyo-style", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/mihoyo-style", "local_weights_removed": true, "bytes_freed": 211734008 }, "path": "ideogram-4/mihoyo-style", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is MiHoYo Style?", "answer": "MiHoYo Style is a style LoRA designed for Ideogram 4.0 checkpoints, published by [allpleoleo439](/users/allpleoleo439) on CivitAI in April 2025, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 55 downloads. It's intended for china, anime, game, and mmorpg content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "china", "style", "anime", "game", "mmorpg" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "allpleoleo439", "url": "https://civarchive.com/models/995865?modelVersionId=3061630" } }, { "name": "Disco Elysium", "slug": "disco-elysium", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2487755, "version_id": 3129598, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/2487755", "repo_id": null, "file_path": null, "author": "wystanbelov785", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Needs a reduced strength; the default 1.0 is likely too strong. [source: description]", "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (251 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "MixedElysium_Ideogram.safetensors", "size_bytes": 211734080, "size_mb": 201.93, "sha256": "e762d7696337f6e50f78a7316a37ef2bcea847ab9a712f5ac5f1fa91697c0caa" }, "stats": { "downloads": 251, "thumbs_up": 22, "thumbs_down": 0 }, "popularity": { "downloads": 251, "thumbs_up": 22, "thumbs_down": 0, "comments": 0, "approval": "100% positive (22/22)", "age_days": 52.5, "downloads_per_day": 0.71, "rank_by_downloads": 26, "rank_by_momentum": 27, "of_total": 35, "adoption_tier": "Modest", "momentum": "Long tail" }, "selection_score": 39.5, "selection_reasons": [ "251 downloads", "22 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "style", "cartoon", "digital art", "oil painting", "illustration" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "disco-elysium", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/disco-elysium", "local_weights_removed": true, "bytes_freed": 211734080 }, "path": "ideogram-4/disco-elysium", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Disco Elysium?", "answer": "Disco Elysium is a style LoRA designed for Ideogram 4.0 checkpoints, published by [wystanbelov785](/users/wystanbelov785) on CivitAI in March 2026, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 37 downloads. It's intended for digital art, oil painting, illustration, and cartoon content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "digital art", "oil painting", "style", "illustration", "cartoon" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "wystanbelov785", "url": "https://civarchive.com/models/2487755?modelVersionId=3129598" } }, { "name": "Gray Screen bypass (LoRA and workflow)", "slug": "gray-screen-bypass-lora-and-workflow", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2750357, "version_id": 3094072, "version_name": "v1.0", "source": "civitai", "url": "https://civitai.com/models/2750357", "repo_id": null, "file_path": null, "author": "wystanbelov785", "category": { "key": "style", "label": "Visual style" }, "modalities": [ "t2i" ], "primary_modality": "t2i", "checkpoint_family": "transformer", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The main 9.3B DiT and what essentially every Ideogram 4 LoRA targets. diffusers' README_ideogram4.md is explicit: \"This trainer adds LoRA to the conditional `transformer` only; the unconditional one stays frozen.\" In ComfyUI this is `ideogram4_fp8_scaled.safetensors` (9.28 GB) or its int8_convrot / nvfp4_mixed equivalents." ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Modest adoption (241 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "Gray_000002000.safetensors", "size_bytes": 211734024, "size_mb": 201.93, "sha256": "0292898e3f3cd3edbb15bcff80e81a81a5f4e5fc221438fc9fd0b5fcab89ae78" }, "stats": { "downloads": 241, "thumbs_up": 22, "thumbs_down": 0 }, "popularity": { "downloads": 241, "thumbs_up": 22, "thumbs_down": 0, "comments": 0, "approval": "100% positive (22/22)", "age_days": 63.8, "downloads_per_day": 3.78, "rank_by_downloads": 27, "rank_by_momentum": 8, "of_total": 35, "adoption_tier": "Niche", "momentum": "Trending" }, "selection_score": 57.0, "selection_reasons": [ "241 downloads", "22 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1824x1248", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "2128x1456", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "2128x1456", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "2432x1664", "prompt": "" } ], "tags": [ "tool" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "gray-screen-bypass-lora-and-workflow", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/gray-screen-bypass-lora-and-workflow", "local_weights_removed": true, "bytes_freed": 211734024 }, "path": "ideogram-4/gray-screen-bypass-lora-and-workflow", "archive": { "comments_total": 3, "comments": [ { "author": "8404959", "score": 0, "text": "Works well", "created_at": "2026-07-04T04:26:14.819Z" }, { "author": "12436275", "score": 0, "text": "image blocked by safety filter again", "created_at": "2026-07-22T21:05:33.126Z" } ], "faq": [ { "question": "What is Gray Screen bypass (LoRA and workflow)?", "answer": "Gray Screen bypass (LoRA and workflow) is a LoRA designed for Ideogram 4.0 checkpoints, published by [wystanbelov785](/users/wystanbelov785) on CivitAI in July 2026. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 241 downloads. It's intended for tool content." } ], "trigger_words": [], "tags": [ "tool" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "wystanbelov785", "url": "https://civarchive.com/models/2750357?modelVersionId=3094072" } }, { "name": "Sorcery & Steel - Ideogram", "slug": "sorcery-steel-ideogram", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2818244, "version_id": 3045842, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/2818244", "repo_id": null, "file_path": null, "author": "ArsMachina", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "Sorcery & Steel," ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Explicit trigger word(s) — Sorcery & Steel, — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (231 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "SorcerySteel_Ideogram4_000006000..safetensors", "size_bytes": 211734016, "size_mb": 201.93, "sha256": "8ad793c5256153c02bdabe77dbabd621e47de25a13b3630d397ae37407b62273" }, "stats": { "downloads": 231, "thumbs_up": 30, "thumbs_down": 0 }, "popularity": { "downloads": 231, "thumbs_up": 30, "thumbs_down": 0, "comments": 0, "approval": "100% positive (30/30)", "age_days": 75.8, "downloads_per_day": 3.05, "rank_by_downloads": 28, "rank_by_momentum": 11, "of_total": 35, "adoption_tier": "Niche", "momentum": "Steady" }, "selection_score": 51.9, "selection_reasons": [ "231 downloads", "30 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "848x1264", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "848x1264", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "848x1264", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "848x1264", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "848x1264", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "848x1264", "prompt": "" } ], "tags": [ "d&d", "90s", "80s", "fantasy", "metal", "comic", "style" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "sorcery-steel-ideogram", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/sorcery-steel-ideogram", "local_weights_removed": true, "bytes_freed": 211734016 }, "path": "ideogram-4/sorcery-steel-ideogram", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Sorcery & Steel - Ideogram?", "answer": "Sorcery & Steel - Ideogram is a style LoRA designed for Ideogram 4.0 checkpoints, published by [ArsMachina](/users/ArsMachina) on CivitAI in June 2026. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 231 downloads. It's intended for d&d, 90s, 80s, fantasy, and metal content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "Sorcery & Steel," ], "tags": [ "d&d", "90s", "80s", "fantasy", "metal", "comic", "style" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "ArsMachina", "url": "https://civarchive.com/models/2818244?modelVersionId=3045842" } }, { "name": "Art The Clown - Terrifier", "slug": "art-the-clown-terrifier", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2763837, "version_id": 3116134, "version_name": "Ideogram4", "source": "civitai", "url": "https://civitai.com/models/2763837", "repo_id": null, "file_path": null, "author": "Plaguekind", "category": { "key": "character", "label": "Character / subject" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Large at 1071 MB — slower to load and heavier to stack. [source: file size]", "Modest adoption (221 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "4rtcl0wn_id4.safetensors", "size_bytes": 1123105152, "size_mb": 1071.08, "sha256": "d6794f4b0c0866d9e6adb5be781c2ebd82ad268d5c4e31b82d5f3710cc0ccb0a" }, "stats": { "downloads": 221, "thumbs_up": 24, "thumbs_down": 0 }, "popularity": { "downloads": 221, "thumbs_up": 24, "thumbs_down": 0, "comments": 0, "approval": "100% positive (24/24)", "age_days": 56.6, "downloads_per_day": 0.79, "rank_by_downloads": 29, "rank_by_momentum": 25, "of_total": 35, "adoption_tier": "Niche", "momentum": "Long tail" }, "selection_score": 38.5, "selection_reasons": [ "221 downloads", "24 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1536x2048", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1344x2368", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1536x2048", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1536x2048", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1248x1664", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1536x2048", "prompt": "" } ], "tags": [ "terrifier", "character", "krea2", "krea", "villian", "horror", "clown" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "art-the-clown-terrifier", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/art-the-clown-terrifier", "local_weights_removed": true, "bytes_freed": 1123105152 }, "path": "ideogram-4/art-the-clown-terrifier", "archive": { "comments_total": 2, "comments": [ { "author": "8469732", "score": 0, "text": "The Ideogram version has some intimidating images but the Krea 2 version seems pretty tame, more of a subtle humor than frightening.", "created_at": "2026-07-13T04:41:40.777Z" } ], "faq": [ { "question": "What is Art The Clown - Terrifier?", "answer": "Art The Clown - Terrifier is a character LoRA designed for Ideogram 4.0 checkpoints, published by [Plaguekind](/users/Plaguekind) on CivitAI in July 2026, now on version 2. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 45 downloads. It's intended for horror, clown, villian, terrifier, and krea content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "character", "horror", "clown", "villian", "terrifier", "krea", "krea2" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "Plaguekind", "url": "https://civarchive.com/models/2763837?modelVersionId=3116134" } }, { "name": "Taiwan bento (packed lunch) | 便當", "slug": "taiwan-bento-packed-lunch", "base_model": "Ideogram 4", "media_type": "image", "model_id": 952849, "version_id": 3052339, "version_name": "v1.0 | ideogram4", "source": "civitai", "url": "https://civitai.com/models/952849", "repo_id": null, "file_path": null, "author": "allpleoleo439", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (196 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "idegoram4_bando.safetensors", "size_bytes": 211734000, "size_mb": 201.93, "sha256": "271f930729f3e3ccf32644faf7ebfaf75a010913605434eed03bc775d1cd4855" }, "stats": { "downloads": 196, "thumbs_up": 21, "thumbs_down": 0 }, "popularity": { "downloads": 196, "thumbs_up": 21, "thumbs_down": 0, "comments": 0, "approval": "100% positive (21/21)", "age_days": 76.7, "downloads_per_day": 0.56, "rank_by_downloads": 30, "rank_by_momentum": 30, "of_total": 35, "adoption_tier": "Niche", "momentum": "Long tail" }, "selection_score": 30.5, "selection_reasons": [ "196 downloads", "21 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "05_image.webp", "type": "image", "resolution": "768x1376", "prompt": "", "source": "civarchive", "civarchive_id": 134318691 } ], "tags": [ "objects", "food", "taiwan" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "taiwan-bento-packed-lunch", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/taiwan-bento-packed-lunch", "local_weights_removed": true, "bytes_freed": 211734000 }, "path": "ideogram-4/taiwan-bento-packed-lunch", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Taiwan bento (packed lunch) | 便當?", "answer": "Taiwan bento (packed lunch) | 便當 is a LoRA designed for Ideogram 4.0 checkpoints, published by [allpleoleo439](/users/allpleoleo439) on CivitAI in April 2025, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 43 downloads. It's intended for objects, food, and taiwan content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "objects", "food", "taiwan" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "allpleoleo439", "url": "https://civarchive.com/models/952849?modelVersionId=3052339" } }, { "name": "Pale Skin / Redhead Enhancer", "slug": "pale-skin-redhead-enhancer", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2738391, "version_id": 3079368, "version_name": "v1.0", "source": "civitai", "url": "https://civitai.com/models/2738391", "repo_id": null, "file_path": null, "author": "CallMeMaybe", "category": { "key": "detail", "label": "Detail / anatomy fixer" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "pale" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Explicit trigger word(s) — pale — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 101 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author says the LoRA may be overtrained — start at low strength. [source: description]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (194 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "pale-id4-v2_000000500.safetensors", "size_bytes": 105892920, "size_mb": 100.99, "sha256": "138d813cb072bc3a5317072507cb112cbf326caf55c846ccabb99dd3e8d22643" }, "stats": { "downloads": 194, "thumbs_up": 11, "thumbs_down": 0 }, "popularity": { "downloads": 194, "thumbs_up": 11, "thumbs_down": 0, "comments": 0, "approval": "100% positive (11/11)", "age_days": 68.3, "downloads_per_day": 2.84, "rank_by_downloads": 31, "rank_by_momentum": 12, "of_total": 35, "adoption_tier": "Niche", "momentum": "Steady" }, "selection_score": 48.9, "selection_reasons": [ "194 downloads", "11 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1264x1584", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "2528x1584", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1264x1584", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1264x1584", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "752x1328", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "752x1328", "prompt": "" } ], "tags": [ "pale skin", "realistic", "skin", "pale", "concept" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "pale-skin-redhead-enhancer", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/pale-skin-redhead-enhancer", "local_weights_removed": true, "bytes_freed": 105892920 }, "path": "ideogram-4/pale-skin-redhead-enhancer", "archive": { "comments_total": 1, "comments": [ { "author": "1547088", "score": 0, "text": "Try lowering steps to 20 during generation and increase reso", "created_at": "2026-06-29T23:54:20.289Z" } ], "faq": [ { "question": "What is Pale Skin / Redhead Enhancer?", "answer": "Pale Skin / Redhead Enhancer is a concept LoRA designed for Ideogram 4.0 checkpoints, published by [CallMeMaybe](/users/CallMeMaybe) on CivitAI in June 2026. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 194 downloads. It's intended for pale skin, realistic, skin, and pale content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "pale" ], "tags": [ "pale skin", "realistic", "skin", "pale", "concept" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "CallMeMaybe", "url": "https://civarchive.com/models/2738391?modelVersionId=3079368" } }, { "name": "Super Robot Style", "slug": "super-robot-style", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1029997, "version_id": 3051798, "version_name": "v1.0 | ideogram4", "source": "civitai", "url": "https://civitai.com/models/1029997", "repo_id": null, "file_path": null, "author": "allpleoleo439", "category": { "key": "camera", "label": "Camera movement" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Needs a reduced strength; the default 1.0 is likely too strong. [source: description]", "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (144 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "idegoram4_ROBT.safetensors", "size_bytes": 211734000, "size_mb": 201.93, "sha256": "97d3f116d163d084553af2083e253560b5a610ad98ef8b47772490aeede2cf9d" }, "stats": { "downloads": 144, "thumbs_up": 21, "thumbs_down": 0 }, "popularity": { "downloads": 144, "thumbs_up": 21, "thumbs_down": 0, "comments": 0, "approval": "100% positive (21/21)", "age_days": 76.9, "downloads_per_day": 0.6, "rank_by_downloads": 32, "rank_by_momentum": 29, "of_total": 35, "adoption_tier": "Niche", "momentum": "Long tail" }, "selection_score": 29.3, "selection_reasons": [ "144 downloads", "21 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1184x896", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1376x768", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1376x768", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "anime", "japanime", "japan", "cartoon", "style", "robot" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "super-robot-style", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/super-robot-style", "local_weights_removed": true, "bytes_freed": 211734000 }, "path": "ideogram-4/super-robot-style", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Super Robot Style?", "answer": "Super Robot Style is a style LoRA designed for Ideogram 4.0 checkpoints, published by [allpleoleo439](/users/allpleoleo439) on CivitAI in April 2025, now on version 2. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 46 downloads. It's intended for anime, japanime, japan, cartoon, and robot content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "anime", "japanime", "japan", "cartoon", "style", "robot" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "allpleoleo439", "url": "https://civarchive.com/models/1029997?modelVersionId=3051798" } }, { "name": "Pagani Zonda R - 2009", "slug": "pagani-zonda-r-2009", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2710899, "version_id": 3053389, "version_name": "v1.0 IDEOGRAM4", "source": "civitai", "url": "https://civitai.com/models/2710899", "repo_id": null, "file_path": null, "author": "mastermerlin", "category": { "key": "accelerator", "label": "Speed / step-reduction" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [ "P4ganiZondaR2009" ], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "One real step-reduction LoRA exists: **TurboTime** by ostris (`ostris/ideogram_4_turbotime_lora`, file `ideogram_4_turbotime_v1.safetensors`, 846,778,576 bytes ≈ 808 MiB; also mirrored on CivitAI under the same filename). The model card: it \"is capable of generating good quality images in as little as 2 steps with no CFG and no unconditional model\", made with an in-progress \"continuious turbo training method\", with updated versions likely. The card's own sample grid is at 8 steps. The nomadoor community workflow describes it as a \"2 to 8 step LoRA\" and runs it at 8 steps, cfg 1.0, strength 0.6", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Explicit trigger word(s) — P4ganiZondaR2009 — so the effect can be turned on and off from the prompt. [source: model metadata]", "Compact at 163 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "Author labels this experimental — results vary shot to shot. [source: description]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (144 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "mastermerlin_ideogram4-fp8_Pagani_Zonda_R_2009_final.safetensors", "size_bytes": 170503408, "size_mb": 162.6, "sha256": "e93e9b2bbf582dbe0e8b744eec81e5fd282f9538eaa2d5f72334cff703a2e6d5" }, "stats": { "downloads": 144, "thumbs_up": 7, "thumbs_down": 0 }, "popularity": { "downloads": 144, "thumbs_up": 7, "thumbs_down": 0, "comments": 0, "approval": "100% positive (7/7)", "age_days": 75.7, "downloads_per_day": 0.41, "rank_by_downloads": 33, "rank_by_momentum": 34, "of_total": 35, "adoption_tier": "Niche", "momentum": "Long tail" }, "selection_score": 24.2, "selection_reasons": [ "144 downloads", "7 thumbs-up", "documented triggers + usage notes", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1600x896", "prompt": "" } ], "tags": [ "pagani", "vehicle", "zonda" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "pagani-zonda-r-2009", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/pagani-zonda-r-2009", "local_weights_removed": true, "bytes_freed": 170503408 }, "path": "ideogram-4/pagani-zonda-r-2009", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Pagani Zonda R - 2009?", "answer": "Pagani Zonda R - 2009 is a LoRA designed for Ideogram 4.0 checkpoints, published by [mastermerlin](/users/mastermerlin) on CivitAI in June 2026, now on version 5. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 31 downloads. It's intended for pagani, vehicle, and zonda content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [ "P4ganiZondaR2009" ], "tags": [ "pagani", "vehicle", "zonda" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "mastermerlin", "url": "https://civarchive.com/models/2710899?modelVersionId=3053389" } }, { "name": "Salomon XT-6 Sneakers", "slug": "salomon-xt-6-sneakers", "base_model": "Ideogram 4", "media_type": "image", "model_id": 1827315, "version_id": 3153436, "version_name": "Ideogram V2.0", "source": "civitai", "url": "https://civitai.com/models/1827315", "repo_id": null, "file_path": null, "author": "feloxxx75", "category": { "key": "clothing", "label": "Clothing / wardrobe" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Thin documentation — expect trial and error. [source: description length]", "Modest adoption (117 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "salomon_id4.safetensors", "size_bytes": 211734000, "size_mb": 201.93, "sha256": "bcee988a63010113f9470d6d38f5213c829db5152ff01a79caffe576bd4b86b4" }, "stats": { "downloads": 117, "thumbs_up": 6, "thumbs_down": 0 }, "popularity": { "downloads": 117, "thumbs_up": 6, "thumbs_down": 0, "comments": 0, "approval": "100% positive (6/6)", "age_days": 44.6, "downloads_per_day": 0.85, "rank_by_downloads": 34, "rank_by_momentum": 24, "of_total": 35, "adoption_tier": "Niche", "momentum": "Long tail · fresh release" }, "selection_score": 36.4, "selection_reasons": [ "117 downloads", "6 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "" }, { "file": "04_image.webp", "type": "image", "resolution": "1088x1936", "prompt": "", "source": "civarchive", "civarchive_id": 137468709 }, { "file": "05_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "", "source": "civarchive", "civarchive_id": 137468715 }, { "file": "06_image.jpeg", "type": "image", "resolution": "1088x1936", "prompt": "", "source": "civarchive", "civarchive_id": 137468733 } ], "tags": [ "sneakers", "salomon", "fashion", "clothing" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "salomon-xt-6-sneakers", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/salomon-xt-6-sneakers", "local_weights_removed": true, "bytes_freed": 211734000 }, "path": "ideogram-4/salomon-xt-6-sneakers", "archive": { "comments_total": 0, "comments": [], "faq": [ { "question": "What is Salomon XT-6 Sneakers?", "answer": "Salomon XT-6 Sneakers is a clothing LoRA designed for Ideogram 4.0 checkpoints, published by [feloxxx75](/users/feloxxx75) on CivitAI in August 2025, now on version 3. Load a compatible checkpoint first, then apply this LoRA on top via your UI's LoRA loader. It has attracted 38 downloads. It's intended for sneakers, salomon, and fashion content. The model is flagged NSFW on CivitAI and can produce explicit content." } ], "trigger_words": [], "tags": [ "sneakers", "salomon", "fashion", "clothing" ], "rating": 0, "rating_count": 0, "favorite_count": 0, "creator": "feloxxx75", "url": "https://civarchive.com/models/1827315?modelVersionId=3153436" } }, { "name": "Dream Loom", "slug": "dream-loom", "base_model": "Ideogram 4", "media_type": "image", "model_id": 2605989, "version_id": 3113138, "version_name": "Ideogram", "source": "civitai", "url": "https://civitai.com/models/2605989", "repo_id": null, "file_path": null, "author": "wystanbelov785", "category": { "key": "style", "label": "Visual style" }, "modalities": [], "primary_modality": "unspecified", "checkpoint_family": "unknown", "trigger_words": [], "recommended_settings": {}, "chaining": { "author": { "stacks": "not stated", "notes": "", "avoid": [], "requires": [], "order": "", "strength": "", "stated": false }, "general_rules": [ "Chain `LoraLoaderModelOnly` nodes in series off the CONDITIONAL model loader, ahead of the guider. DeverStyle states its samples were made \"by setting the LORA @ 0.6 strength on the model only, not on the model_negative\".", "Model-only loaders throughout. The text encoder is a frozen Qwen3-VL-8B with no LoRA-patchable CLIP branch, so the model+clip `LoraLoader` has nothing useful to do.", "To affect the unconditional/negative branch you must load a second, physically distinct model file and patch that one. Passing the same patched model object into both DualModelGuider inputs is a verified no-op (Comfy-Org issue #14521; kijai confirmed it is expected behaviour).", "Trigger words go INSIDE the JSON caption, in the semantically matching field, not prepended to a prose string. DeverStyle instructs putting `dvr_arch_tv` in `style_description.art_style` (and in the content field); the Kiki LoRA card describes using `art_style: \"Studio Ghibli style\"` with `medium: \"anime_screencap\"`.", "Two versions of the same LoRA can be stacked at reduced strengths: zjourney's author recommends v1 @ 0.35 + v2 @ 0.45 over either alone at full strength.", "TurboTime changes the graph, not just the strength: it runs at cfg 1.0 and the reference community workflow swaps `DualModelGuider` for a single-model `CFGGuider`, dropping the unconditional model. Re-tune any stacked style LoRA downward.", "A LoRA can be applied to only part of the run. The Gray Screen bypass LoRA is used at NEGATIVE weight (about -0.25) and, in its shipped workflows, only on the first step and then disabled; its author warns that \"LoRA being active through the whole generation could lead to degradation\".", "The author never states which weight family this targets. Silent degradation, not hard errors, in every case observed. (1) Loading a conditional-trained LoRA onto the unconditional transformer succeeds silently because the architectures match; it perturbs the negative branch instead of the positive one. (2) Verified ComfyUI trap: in `Du" ] }, "pros": [ "Compact at 202 MB — cheap to stack with other LoRAs. [source: file size]" ], "cons": [ "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]", "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]", "Modest adoption (89 downloads) — less community feedback to rely on. [source: source stats]" ], "file": { "name": "Sime2_Ideogram.safetensors", "size_bytes": 211734048, "size_mb": 201.93, "sha256": "fdf61f3f7f4bb948c30c5ed01e87e0b14fbee745c3cfac7dd0944833ee98f16e" }, "stats": { "downloads": 89, "thumbs_up": 4, "thumbs_down": 0 }, "popularity": { "downloads": 89, "thumbs_up": 4, "thumbs_down": 0, "comments": 0, "approval": "100% positive (4/4)", "age_days": 57.6, "downloads_per_day": 0.64, "rank_by_downloads": 35, "rank_by_momentum": 28, "of_total": 35, "adoption_tier": "Niche", "momentum": "Long tail" }, "selection_score": 30.0, "selection_reasons": [ "89 downloads", "4 thumbs-up", "top-ranked for its category" ], "example_media": [ { "file": "01_image.jpeg", "type": "image", "resolution": "768x1376", "prompt": "" }, { "file": "02_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "03_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "04_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "05_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" }, { "file": "06_image.jpeg", "type": "image", "resolution": "1024x1024", "prompt": "" } ], "tags": [ "illustration", "style", "watercolor", "oil painting" ], "scraped_at": "2026-09-05T03:22:23+00:00", "mirror": { "repo_id": "Alex995647/loras-ideogram-4", "path_in_repo": "dream-loom", "url": "https://huggingface.co/Alex995647/loras-ideogram-4/tree/main/dream-loom", "local_weights_removed": true, "bytes_freed": 211734048 }, "path": "ideogram-4/dream-loom" } ], "skipped": [] }