Upload folder using huggingface_hub

guides/image/flux-1-dev.md

@@ -0,0 +1,64 @@
+---
+id: flux-1-dev
+displayName: Flux.1 Dev
+aliases:
+  - flux 1 dev
+  - flux.1 dev
+  - flux dev
+  - flux one dev
+modality: image
+tasks:
+  - text-to-image
+  - image-editing
+summary: Natural-language Flux.1 Dev prompt guidance for cinematic image generation, lighting, materials, color, and positive constraint phrasing.
+---
+
+# Flux.1 Dev Prompt Guide
+
+## Best For
+
+- Rich image prompts written as natural cinematic prose.
+- Scenes that need strong lighting, reflections, material behavior, atmosphere, or surreal styling.
+- Positive constraint phrasing instead of native negative prompts.
+
+## Prompt Shape
+
+```text
+POSITIVE PROMPT:
+[main subject and action]. [environment and spatial layout]. [lighting source, direction, color, glow, reflections]. [atmosphere and mood]. [composition and framing]. [material and texture details]. masterpiece, best quality, ultra-detailed, sharp focus, vibrant colors, high contrast, cinematic lighting, flawless composition.
+
+POSITIVE AVOIDANCE PHRASING:
+[clean anatomy, sharp focus, clean edges, balanced exposure, no unwanted text or watermarks phrased as desired outcomes]
+
+ADDITIONAL PARAMETERS:
+Guidance Scale: 3.5
+Steps: 20-50
+Aspect Ratio: [3:4, 9:16, or 16:9]
+Style Strength: [High artistic or Balanced]
+```
+
+## Key Rules
+
+- Lead with the main subject. Word order matters.
+- Use vivid present-tense prose, not rigid token lists.
+- Preserve the user's characters, wardrobe, props, actions, colors, and spatial relationships.
+- Avoid native negative prompts. Reinforce desired outcomes positively.
+- Describe lighting quality, color temperature, glow, reflections, bloom, rim light, and material interaction.
+- Use precise spatial language: foreground, background, centered, symmetrical, leading lines, depth.
+- Add useful material details: fabric texture, individual hair strands, wet gloss, worn metal, surface reflections.
+
+## Strong Prompt Example
+
+```text
+POSITIVE PROMPT:
+A young woman sits on the edge of a low bed in a cozy wood-paneled bedroom at night, gripping a black game controller with both hands while leaning forward with a triumphant playful smile. The room is arranged with green bedding in the foreground and a softly blurred dual-monitor gaming desk in the background. A circular amber wall light above the bed casts warm neon rim light across her face, hair, controller, and shirt graphic, with soft bloom on the wood panels and gentle reflections on the controller's glossy plastic. The mood is intimate, cinematic, and focused, framed as an eye-level medium shot with shallow depth of field and realistic skin texture, fabric weave, natural hair detail, atmospheric shadows, masterpiece, best quality, ultra-detailed, sharp focus, vibrant warm color, high contrast, cinematic lighting, flawless composition.
+
+POSITIVE AVOIDANCE PHRASING:
+clean natural anatomy, accurate hands on the controller, sharp facial focus, balanced exposure, clean image without unwanted text or watermarks, coherent room geometry.
+
+ADDITIONAL PARAMETERS:
+Guidance Scale: 3.5
+Steps: 20-50
+Aspect Ratio: 4:3
+Style Strength: Balanced
+```

guides/image/flux-2-klein.md

@@ -0,0 +1,64 @@
+---
+id: flux-2-klein
+displayName: Flux.2 Klein
+aliases:
+  - flux 2 klein
+  - flux.2 klein
+  - flux2 klein
+  - flux klein
+modality: image
+tasks:
+  - text-to-image
+  - image-editing
+summary: Fast flowing-prose prompt guidance for FLUX.2 Klein 4B and 9B variants, including reference-image roles and low-step settings.
+---
+
+# Flux.2 Klein Prompt Guide
+
+## Best For
+
+- Fast sub-second or low-step image generation on consumer hardware.
+- Flowing prose prompts for text-to-image and image editing with up to four references.
+- Neon, glow, reflections, volumetric lighting, surreal scenes, and material interactions.
+
+## Prompt Shape
+
+```text
+POSITIVE PROMPT:
+[flowing prose: subject first, environment, lighting/color interactions, sensory material details, atmosphere, composition, quality]
+
+POSITIVE AVOIDANCE PHRASING:
+[concise positive guidance for anatomy, focus, details, exposure, glow, reflections, and clean composition]
+
+ADDITIONAL PARAMETERS:
+Steps: 4-8 distilled or 20-50 base
+Guidance Scale: 1.0-3.5
+Aspect Ratio: [3:4, 9:16, or 16:9]
+Variant: [4B for speed or 9B for higher detail]
+```
+
+## Key Rules
+
+- Write connected prose paragraphs, not token lists.
+- Lead with the main subject and exact view.
+- Preserve the user's stated characters, wardrobe, props, actions, colors, and spatial relationships.
+- For reference images, assign roles clearly: composition, subject, wardrobe, palette, material, lighting, or mood.
+- Describe lighting source, direction, color temperature, glow, bloom, reflections, and material interaction.
+- Use positive phrasing instead of native negative prompts.
+- Include at least one useful texture, one reflective quality, and an atmospheric feel when relevant.
+
+## Strong Prompt Example
+
+```text
+POSITIVE PROMPT:
+A young woman with long slightly wavy brown hair sits at the edge of a low bed in a cozy nighttime gamer bedroom, holding a sleek black controller in both hands while leaning forward with a confident playful grin. The green bedding, wood-paneled wall, circular amber wall light, string lights, and softly blurred dual-monitor desk create a warm layered room behind her. The circular light glows at roughly 2700K, spilling yellow-orange bloom across the wood grain, catching subtle reflections on the controller plastic, and rim-lighting strands of hair around her face. The shirt fabric reads as soft dark olive cotton with a clear retro sunset and pine graphic, paired with dark loose cargo pants. The atmosphere feels intimate, competitive, and cinematic, composed as a steady eye-level medium shot with shallow depth of field, sharp facial focus, high contrast, vibrant warm glow, ultra-detailed realistic texture, flawless composition.
+
+POSITIVE AVOIDANCE PHRASING:
+accurate hands and controller grip, natural anatomy, crisp facial detail, balanced warm exposure, clean image without stray text or watermark, coherent bedroom layout.
+
+ADDITIONAL PARAMETERS:
+Steps: 4-8 distilled
+Guidance Scale: 2.0
+Aspect Ratio: 4:3
+Variant: 9B for higher detail
+```

guides/image/z-image-turbo.md

@@ -0,0 +1,61 @@
+---
+id: z-image-turbo
+displayName: Z-Image Turbo
+aliases:
+  - z image turbo
+  - z-image turbo
+  - zimage turbo
+  - z image
+modality: image
+tasks:
+  - text-to-image
+  - image-editing
+summary: Positive-only Z-Image Turbo prompt guidance for fast few-step image generation, text rendering, and safety/quality constraints.
+---
+
+# Z-Image Turbo Prompt Guide
+
+## Best For
+
+- Fast few-step image generation with strong instruction following.
+- Positive-prompt constraint control where negative prompts are ignored by the official pipeline.
+- Bilingual English/Chinese text rendering, clean product shots, portraits, posters, and controlled style prompts.
+
+## Prompt Shape
+
+```text
+POSITIVE PROMPT:
+[shot and subject]. [adult role or functional label if human]. [appearance, pose, clothing/coverage]. [environment/background]. [lighting]. [mood]. [style/medium]. [technical quality]. [cleanup constraints as positive text instructions].
+
+ADDITIONAL PARAMETERS:
+Guidance Scale: 0.0
+Steps: 8-12
+Resolution: 1024x1024, 768x1344, or 1344x768
+Seed: fix during iteration, randomize for exploration
+Negative Prompt: ignored by the official pipeline; express constraints positively
+```
+
+## Key Rules
+
+- Put every important instruction in the positive prompt.
+- Use a clear 80-250 word structured prompt when the idea supports it.
+- Preserve exact subjects, actions, colors, clothing, props, spatial relationships, and requested visible text.
+- Define clothing and coverage explicitly for human subjects when ordinary, professional, public, training-data, or safe-for-work imagery is requested.
+- Use concrete camera and lighting language: shot type, angle, lens feel, soft diffused daylight, noir lighting, studio portrait lighting, rim lighting.
+- Put cleanup constraints near the end: no text, no watermark, no logos, natural hands and fingers, correct anatomy, sharp focus, simple background.
+- For requested visible text, wrap the exact words in quotes and specify placement and style.
+- Avoid long negative prompt blocks, CFG assumptions, and Stable Diffusion weight syntax.
+
+## Strong Prompt Example
+
+```text
+POSITIVE PROMPT:
+A medium-shot realistic photograph of an adult software designer seated at a clean desk in a minimal home studio, facing slightly toward the camera with a calm focused expression. The subject wears a dark crewneck sweater and simple trousers, fully clothed and professional, with natural posture and hands resting near a laptop. The background is uncluttered with a plain wall, a small plant, and soft blurred shelves. Soft diffused daylight enters from the left, creating gentle shadows and clear skin texture without harsh contrast. The mood is quiet, thoughtful, and safe for work. Realistic photography, 50mm lens look, shallow depth of field, clean detailed image, natural hands and fingers, correct human anatomy, no logos, no watermark, no extra text, balanced exposure.
+
+ADDITIONAL PARAMETERS:
+Guidance Scale: 0.0
+Steps: 8-12
+Resolution: 1024x1024
+Seed: fix during iteration
+Negative Prompt: ignored by the official pipeline; express constraints positively
+```

guides/video/ltx-2.3.md

@@ -0,0 +1,80 @@
+---
+id: ltx-2.3
+displayName: LTX-2.3
+aliases:
+  - ltx
+  - ltx 2.3
+  - ltx two point three
+  - ltx video
+modality: video
+tasks:
+  - text-to-video
+  - image-to-video
+summary: Cinematic Prompt Relay guidance for LTX-2.3 video prompts with timecoded scenes, camera motion, sound, and temporal consistency.
+---
+
+# LTX-2.3 Prompt Guide
+
+## Best For
+
+- Detailed image-to-video or text-to-video prompts with strong temporal control.
+- Cinematic scenes that need camera movement, subject motion, sound, and continuity.
+- Prompt Relay style output with global persistent traits and timecoded scene changes.
+
+## Prompt Shape
+
+```text
+POSITIVE PROMPT:
+Global Prompt:
+[persistent style, subject identity, reference-image adherence, lighting, lens, quality, sound bed]
+
+Scene 1 [00:00-00:05]:
+[initial static state, spatial blocking, first action, camera, sound]
+
+Scene 2 [00:05-00:10]:
+[only what changes: motion, expression, camera, physics, sound, dialogue]
+
+NEGATIVE PROMPT:
+[temporal, anatomy, watermark, flicker, jitter, and artifact controls]
+
+FPS:
+24
+
+LORA TRIGGERS APPLIED:
+[exact triggers or none]
+```
+
+## Key Rules
+
+- Use present tense only.
+- Preserve exact user keywords, clothing, props, actions, names, and LoRA triggers.
+- If a first frame or reference image is supplied, anchor identity, wardrobe, pose, lighting, props, and environment to it.
+- Always include a specific lens and aperture, natural motion blur, and 24 fps with a 1/48 shutter equivalent.
+- Sound is mandatory. Describe physical sound events and their timing.
+- Use precise spatial blocking: foreground, background, left, right, distance, facing direction.
+- Avoid high-frequency clothing or background patterns unless required.
+- Do not use weight syntax such as `(word:1.2)` or `[word]`.
+- Do not write fade-outs, final summaries, or poetic endings.
+
+## Strong Prompt Example
+
+```text
+POSITIVE PROMPT:
+Global Prompt:
+A cozy cinematic gamer-bedroom scene at night, faithful to the reference image for subject, wardrobe, controller, bed, wall light, and desk layout. Shot on an 85mm f/1.8 lens with shallow depth of field, natural motion blur, 24 fps with a 1/48 shutter equivalent, warm amber neon ring light casting highlights across the subject's face and hair, soft controller button clicks and a low game-audio bed in the room.
+
+Scene 1 [00:00-00:05]:
+The woman sits on the edge of the low green bed in the foreground, facing camera at eye level, both hands gripping the black controller. The gaming desk stays softly blurred in the back right. Her thumbs tap quickly as her jaw tightens and the neon ring hums faintly.
+
+Scene 2 [00:05-00:10]:
+The camera makes a slow controlled push-in as she breaks into a triumphant laugh, leans forward, and gives a smug playful grin toward an unseen opponent. Controller clicks become faster for a beat, then settle under the warm room tone.
+
+NEGATIVE PROMPT:
+watermark, text, signature, duplicate, static, no motion, frozen, bad anatomy, deformed hands, extra limbs, micro jitter, flickering, strobing, aliasing, high frequency patterns, motion artifacts, temporal inconsistency, frame stuttering
+
+FPS:
+24
+
+LORA TRIGGERS APPLIED:
+none
+```

manifest.json

@@ -1,6 +1,44 @@
 {
-  "generatedAt": "2026-07-05T05:05:09Z",
+  "generatedAt": "2026-07-05T05:12:18Z",
   "guides": [
+    {
+      "aliases": [
+        "flux 1 dev",
+        "flux.1 dev",
+        "flux dev",
+        "flux one dev"
+      ],
+      "bytes": 3074,
+      "displayName": "Flux.1 Dev",
+      "id": "flux-1-dev",
+      "modality": "image",
+      "path": "guides/image/flux-1-dev.md",
+      "sha256": "6db8c819358e9a324879cff5fb78c57b69cdc2f26c82f2028038381484899126",
+      "summary": "Natural-language Flux.1 Dev prompt guidance for cinematic image generation, lighting, materials, color, and positive constraint phrasing.",
+      "tasks": [
+        "text-to-image",
+        "image-editing"
+      ]
+    },
+    {
+      "aliases": [
+        "flux 2 klein",
+        "flux.2 klein",
+        "flux2 klein",
+        "flux klein"
+      ],
+      "bytes": 3005,
+      "displayName": "Flux.2 Klein",
+      "id": "flux-2-klein",
+      "modality": "image",
+      "path": "guides/image/flux-2-klein.md",
+      "sha256": "0ffe5bd85e6f17a98d671f6d500dd4edb795d760190f1faf240dd681fa1dfe43",
+      "summary": "Fast flowing-prose prompt guidance for FLUX.2 Klein 4B and 9B variants, including reference-image roles and low-step settings.",
+      "tasks": [
+        "text-to-image",
+        "image-editing"
+      ]
+    },
     {
       "aliases": [
         "krea",
@@ -21,6 +59,25 @@
         "image-editing"
       ]
     },
+    {
+      "aliases": [
+        "z image turbo",
+        "z-image turbo",
+        "zimage turbo",
+        "z image"
+      ],
+      "bytes": 3085,
+      "displayName": "Z-Image Turbo",
+      "id": "z-image-turbo",
+      "modality": "image",
+      "path": "guides/image/z-image-turbo.md",
+      "sha256": "b4944d2a0b308318f9e8d9f463896037f43d663cb1180a6b93ec7a9362022e5d",
+      "summary": "Positive-only Z-Image Turbo prompt guidance for fast few-step image generation, text rendering, and safety/quality constraints.",
+      "tasks": [
+        "text-to-image",
+        "image-editing"
+      ]
+    },
     {
       "aliases": [
         "ace-step",
@@ -57,6 +114,25 @@
         "text-to-music"
       ]
     },
+    {
+      "aliases": [
+        "ltx",
+        "ltx 2.3",
+        "ltx two point three",
+        "ltx video"
+      ],
+      "bytes": 3152,
+      "displayName": "LTX-2.3",
+      "id": "ltx-2.3",
+      "modality": "video",
+      "path": "guides/video/ltx-2.3.md",
+      "sha256": "1404c59c16453fabb4d73a15e635eadf87dd0c6cf8d2573a85c99d3d7e4ca006",
+      "summary": "Cinematic Prompt Relay guidance for LTX-2.3 video prompts with timecoded scenes, camera motion, sound, and temporal consistency.",
+      "tasks": [
+        "text-to-video",
+        "image-to-video"
+      ]
+    },
     {
       "aliases": [
         "seedance",