πΌοΈ ERNIE-AIO
All-in-one ERNIE-Image checkpoints for ComfyUI β Base and Turbo, BF16 and FP8, text encoder and VAE included
ComfyUI/models/checkpoints/, load it with Load Checkpoint (no extra VAE or text-encoder loader needed) and generate. Base: 50 steps, CFG 4.0. Turbo: 8 steps, CFG 1.0.
β¨ At a glance
| Property | Details |
|---|---|
| π§© Type | All-in-one checkpoints with integrated text encoder and VAE |
| ποΈ Base models | baidu/ERNIE-Image and baidu/ERNIE-Image-Turbo |
| π¦ Versions | Base and Turbo, each in BF16 and FP8 |
| ποΈ Settings | Base: 50 steps Β· CFG 4.0 β Turbo: 8 steps Β· CFG 1.0 |
| ποΈ Strengths | Instruction following, text rendering, structured layouts |
| π§° Workflow | ComfyUI workflow with upscaler in Workflow/ |
π¦ Models
| Model | Style | FP8 | BF16 | Best for |
|---|---|---|---|---|
| πΈ ERNIE-AIO-Base | Realistic / structured | β | β | High-quality, refined output |
| β‘ ERNIE-AIO-Turbo | Realistic / structured | β | β | Fast iteration, 8-step generation |
All models include the text encoder and VAE in a single checkpoint, so no extra loaders are needed.
πΈ ERNIE-AIO-Base
The Base version is meant for high-quality, deliberate and stable generation.
Features: π strong instruction following Β· ποΈ text rendering Β· π structured layouts
Steps: 50 | CFG: 4.0
Use this version when you want stronger refinement and more careful rendering, full non-turbo quality, or the best results for posters, infographics and text-heavy layouts.
β‘ ERNIE-AIO-Turbo
The Turbo version is a distilled variant for fast generation and quick iteration.
Features: β‘ 8-step generation Β· π― high prompt fidelity Β· π quick turnaround
Steps: 8 | CFG: 1.0
Use this version when you want fast output for prototyping and iteration, strong visual quality with few steps, or efficient VRAM usage.
π FP8 or BF16?
| π‘ FP8 | π BF16 | |
|---|---|---|
| Quality | about 95% | 100% |
| File size | Smaller | Full |
| Speed | Faster | Standard |
| VAE | FP32 (stable) | FP32 (stable) |
| Best for | Daily use | Final renders |
Note: I quantized the FP8 versions in this repository myself. The VAE is kept at FP32 in all variants for stable decoding.
π§© Workflow
| File | Description |
|---|---|
Workflow/ERNIE-AIO-Upscale.json |
Standard ComfyUI workflow with upscaler |
A ready-to-use ComfyUI workflow is included in the Workflow/ folder. Optional Prompt Enhancer files are included in the prompt-enhancer/ folder.
Minimal workflow: Checkpoint Loader β CLIP Text Encode (Prompt) β Empty Latent Image β KSampler β VAE Decode β Save Image.
ποΈ Prompt Enhancer (optional)
The Prompt Enhancer expands short prompts into richer, more descriptive ones. The included workflow uses it, but everything also works without it.
β οΈ Placement: put the Prompt Enhancer files directly into
ComfyUI/models/text_encoders/, not into aprompt-enhancer/subfolder.
π‘ Prompting
Natural language works best. These models excel at instruction following and text rendering.
β
A cinematic movie poster with a bold readable title at the top,
dramatic lighting, a futuristic skyline in the background,
and a heroic figure silhouetted against a neon-lit city.
β movie poster, city, hero, cinematic, dramatic
ERNIE-Image is especially strong for:
| Category | Examples |
|---|---|
| π¬ Posters and title cards | Movie posters, event flyers, title images |
| β Product ads | Coffee brands, packaging with readable text |
| π Infographics | Labeled sections, structured layouts |
| π¬ Comics | Speech bubbles, panel compositions |
| π₯οΈ UI-style visuals | Interface mockups, app screens |
| πΌοΈ Realistic imagery | Portraits, environments, product shots |
For the best results, describe: the subject and environment Β· the composition and layout Β· the mood and lighting Β· the camera framing Β· any text that should appear in the image.
π₯ Downloads
All files are available directly in this repository. The Civitai versions are:
| Model | Precision | Civitai |
|---|---|---|
| β‘ ERNIE-AIO-Turbo | π BF16 and π‘ FP8 | Download |
| πΈ ERNIE-AIO-Base | π BF16 and π‘ FP8 | Download |
π Files
βββ ERNIE-AIO-Base-bf16.safetensors # πΈ Base β BF16
βββ ERNIE-AIO-Base-fp8.safetensors # πΈ Base β FP8
βββ ERNIE-AIO-Turbo-bf16.safetensors # β‘ Turbo β BF16
βββ ERNIE-AIO-Turbo-fp8.safetensors # β‘ Turbo β FP8
β
βββ Workflow/
β βββ ERNIE-AIO-Upscale.json # π ComfyUI workflow with upscaler
β
βββ prompt-enhancer/ # ποΈ Optional Prompt Enhancer files
β
βββ images/
βββ ernie-aio-base.png # πΈ Base preview image
βββ ernie-aio-turbo.png # β‘ Turbo preview image
π Credits
Original model: Baidu ERNIE-Image team (single-stream DiT, Apache-2.0). AIO conversion and FP8 quantization: SeeSee.
| Resource | Link |
|---|---|
| ERNIE-Image (Hugging Face) | baidu/ERNIE-Image |
| ERNIE-Image-Turbo (Hugging Face) | baidu/ERNIE-Image-Turbo |
| GitHub | baidu/ERNIE-Image |
Download β Load Checkpoint β Generate. π
π§ More from SeeSee β all models
π§© Checkpoints & AIO
| Model | Hugging Face | Civitai |
|---|---|---|
| AniSee | SeeSee21/AniSee | Model Β· Workflows |
| Z-Anime | SeeSee21/Z-Anime | Model |
| Z-Image-Turbo-AIO | SeeSee21/Z-Image-Turbo-AIO | Model Β· Workflow |
| ERNIE-AIO | SeeSee21/ERNIE-AIO | Model Β· Workflow |
| Chroma-Anime-AIO | SeeSee21/Chroma-Anime-AIO | Model Β· Workflow |
| FLUX.2-klein-4B-AIO | SeeSee21/FLUX.2-klein-4B-AIO | Model Β· Workflow |
| QWEN-Anime | SeeSee21/QWEN-Anime | Model Β· Workflow |
| SeeMoss-Genesis | SeeSee21/SeeMoss-Genesis | Model Β· Workflow |
| See-Krea2 | SeeSee21/See-Krea2-Genesis | Model Β· Workflow |
| LTX See Motion | SeeSee21/LTX-See-Motion | Model Β· Workflow |
π LoRAs
| Model | Hugging Face | Civitai |
|---|---|---|
| Illustrious Anime Collection | SeeSee21/Illustrious-Anime-Collection | Model |
| Chroma Semi-Anime | SeeSee21/Chroma-Semi-Anime | Model |
| Kitsune Furry Girl | SeeSee21/Kitsune-Furry-Girl-Z-Image | Model |
| Ayase Momo (Dandadan) | SeeSee21/Ayase-Momo-Z-Image | Model |
π§© Checkpoints & AIO Β· π LoRAs Β· π Civitai Β· βΆοΈ YouTube
Made with β€οΈ by SeeSee
- Downloads last month
- 114
Model tree for SeeSee21/ERNIE-AIO
Base model
baidu/ERNIE-Image
