comfy-ernie-image-turbo

Consolidated model bundle for serverless ERNIE-Image Turbo text-heavy / infographic image generation on RunPod (and any ComfyUI deployment).

What this repo is

A RunPod-Cached-Models-shaped consolidation of the ERNIE-Image Turbo file set into one HF repo, arranged to mirror ComfyUI's models/<subdir>/ tree โ€” so a serverless endpoint (or a local container) can pull everything in a single warmup pass. All weights are copied unchanged from the upstream Comfy-Org/ERNIE-Image repo (turbo variant only; the full 16 GB ernie-image.safetensors is intentionally excluded).

Files

Path ~Size Upstream
diffusion_models/ernie-image-turbo.safetensors 16.1 GB Comfy-Org/ERNIE-Image
text_encoders/ministral-3-3b.safetensors 7.7 GB Comfy-Org/ERNIE-Image
text_encoders/ernie-image-prompt-enhancer.safetensors 6.9 GB Comfy-Org/ERNIE-Image
vae/flux2-vae.safetensors 0.3 GB Comfy-Org/ERNIE-Image

Total: ~31 GB.

Runtime layout

comfy_paths.json declares the mapping from this repo's dirs to ComfyUI's models/<subdir>/ tree. The container boot script (symlink-models.sh) reads this manifest and creates symlinks โ€” no copies, no path inference.

Workflow

workflow.json is the canonical API-format ComfyUI workflow this endpoint serves (native ERNIE-Image Turbo template; prompt-enhancement disabled by default so literal English prompts reach the model). Native ComfyUI nodes โ€” no custom nodes required.

License

ERNIE-Image weights are governed by the upstream model license (Baidu / Comfy-Org). Use of this consolidated repo is subject to the same terms.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support