Comfy-Org Makes Qwen-Image Work in ComfyUI Without Complex Setup
Comfy-Org maintains drop-in single-file safetensors for Alibaba's Qwen-Image family, bundling the base model, the 2512 refresh, and a distilled variant for ComfyUI.
- Comfy-Org publishes repackaged single-file safetensors of Alibaba's Qwen-Image for drop-in ComfyUI use.
- Bundle includes the original Qwen-Image, the Qwen-Image-2512 refresh, and the DiffSynth distilled variant.
- Multiple precisions ship side by side: bf16, fp8_e4m3fn, fp8 mixed, fp8 hq, and nvfp4 for Blackwell GPUs.
- Matching Qwen2.5-VL 7B text encoder and Qwen-Image VAE are included so pairings stay consistent.
- Files slot into standard
diffusion_models,text_encoders, andvaefolders with no diffusers conversion. - Separate Comfy-Org/Qwen-Image-2.1 repo hosts the newer 7B RGBA editing model.
Installing Qwen-Image from its upstream Hugging Face repositories requires developers to reconcile Diffusers shards, configuration files, a VAE, and the correct text encoder. The Comfy-Org repository converts those components into individual safetensors files that fit ComfyUI’s standard model directories.
The package draws from the original Qwen-Image release, the 2512 refresh, and DiffSynth-Studio’s distilled checkpoint. Comfy-Org maintains the packaging, while the architectures and weights come from those upstream projects. Their licenses and model cards continue to govern use.
Three folders hold the stack
Each component performs a separate job, so a working workflow generally needs one diffusion model, one text encoder, and the shared VAE.
| Component | Purpose | ComfyUI destination |
|---|---|---|
| Diffusion model | Turns noisy latent data into the generated image | ComfyUI/models/diffusion_models/ |
| Text encoder | Converts prompts into conditioning data for the model | ComfyUI/models/text_encoders/ |
| VAE | Converts between image pixels and latent representations | ComfyUI/models/vae/ |
The diffusion directory includes several model generations and precision levels:
qwen_image_bf16.safetensorsandqwen_image_fp8_e4m3fn.safetensorspackage the original Qwen-Image checkpoint.qwen_image_2512_bf16.safetensorsand its FP8 counterpart package the Qwen-Image-2512 refresh.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
