Qwen Image 2.1 fix v2.0 (updated due to feedback), plus a new sampler
retained text
This is actually two models: **Qwen Image 2.1 Fix v2.0** and **Qwen Image 2.1 Fix Opinionated v1.0**. The former surgically targets Qwen's weird noise and messy details with almost no effect on composition, and the latter in my opinion produces higher quality images but alters composition a bit to do so. In either case, you'll get images that are similar to Qwen Image 2.1's default output, but less noisy and with cleaner details. Download links: * [https://civitai.com/models/2957332/qwen-image-21-fix-v20?modelVersionId=3379469](https://civitai.com/models/2957332/qwen-image-21-fix-v20?modelVersionId=3379469) * [https://huggingface.co/e-n-v-y/Qwen-Image-2.1-Fix-v2.0](https://huggingface.co/e-n-v-y/Qwen-Image-2.1-Fix-v2.0) * [https://huggingface.co/e-n-v-y/Qwen-Image-2.1-Fix-Opinionated-v1.0](https://huggingface.co/e-n-v-y/Qwen-Image-2.1-Fix-Opinionated-v1.0) Also grab the res\_2m\_nc sampler (modified from RES4LYF's res\_2m) and a number of other samplers customized for Qwen 2.1 here: [https://github.com/envy-ai/ComfyUI-DPMpp-2M-Sharp](https://github.com/envy-ai/ComfyUI-DPMpp-2M-Sharp) (you can also find this on the Comfy registry) The workflow I used to produce these images (along with the prompts) can be found at the download links.
document fetched 2026-10-04 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1wwi34c/qwen_image_21_fix_v20_updated_due_to_feedback