lllyasviel / lllyasviel/stable-diffusion-webui-forge

img2img saturation loss

Open
#1,509 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
13k
Forks
1.7k
PR merge metrics
No merged PRs in 30d

Description

After install of Forge UI (Forge with CUDA 12.1 + Pytorch 2.3.1, I did an update and then a full reinstall), I'm experiencing a loss of saturation in generated images every time I use inpainting or img2img upscale (SD Upscale).
I'm using photo-realistic SDXL models (but a quick test with FLux gave the same result of desaturated image): I tried to assign a VAE or leaving it in auto, and also set the "Apply color correction to img2img results to match original colors" but without changes, the images saved are always a desaturated version of the orignal. This does not happen if using HiRes Fix.

OS: Windows
Browser: Firefox
Graphic: RTX2060 super 8GB

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the saturation loss with inpainting and img2img upscale using the reported Forge, CUDA, PyTorch, model, and Windows setup, then compare the saved output with HiRes Fix. Trace the paths responsible for img2img color handling, VAE selection, and the color-correction option; done means these workflows preserve the original image's saturation.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.