Diffusion Single File
comfyui

I fixed the "Yellow Tint" issue

#10
by K0KU20 - opened

I've made a custom node called "ComfyUI-Bleachery" that will fix Qwen-Image-2.1 "Yellow Tint" issue. Just simply connect this node between your image output node and save image node. That's all.

It also supports batch process, so you can do for lots of images at one time, and also supports transparency images/masks as well. Please check the before/after result from below images I attached.

You can get "ComfyUI-Bleachery" from my repo.
https://github.com/K0KU20/ComfyUI-Bleachery

pythonw_5eGLuzbPbn

Before-Normal (Yellow Tint)Qwen_image_2.1_00002

After-Normal (White balanced)Qwen_image_2.1_00003

Before-Transparency (Yellow Tint)Qwen_image_2.1_00005

After-Transparency (White balanced)Qwen_image_2.1_00006

其实你只要提高CFG,并加入负面提示词就可以解决了

其实你只要提高CFG,并加入负面提示词就可以解决了

Totally agree. But I think my solution might be a little faster.
The inference time will become longer when you raise CFG and input negative prompt.

其实你只要提高CFG,并加入负面提示词就可以解决了

Totally agree. But I think my solution might be a little faster.
The inference time will become longer when you raise CFG and input negative prompt.

确实会增加时长,但负面提示还可以解决手指畸形、画面昏暗、多肢体等其它问题。

Sign up or log in to comment