search

AUTOMATIC1111 / stable-diffusion-webui

Stable Diffusion web UI
GNU Affero General Public License v3.0
135.13k stars 25.81k forks source link

[Feature Request]: Add option for postprocessing accordion #13114

Open narukaze132 opened 10 months ago

narukaze132 commented 10 months ago

Is there an existing issue for this?

What would your feature do ?

Currently, it's possible to add "postprocessing" to normal txt2img and img2img operations, allowing you to set up a workflow to immediately upscale or apply GFPGAN or CodeFormer to the result, rather than having to send it to the Extras tab and do it there. However, as it stands, those options will always appear in the main UI, with no option to roll them up into an accordion so you don't have them occupying UI space when you're not planning to use them. I would like it if I could hide those options when I'm not planning to use them.

I'm not sure how to describe this in a concise manner, but you know how you can add options (like clip-skip, as one example I use), to those tabs, and you can set it up so you can roll those up into an accordion when you're not using them? I'd like something like that, but for the postprocessing stuff.

Proposed workflow

First off, to enable the setting:

  1. Go to the Postprocessing menu in Settings (or User interface, I'm not picky)
  2. Make sure some of the Postprocessing operations are set up to appear in the UI
  3. Check off a checkbox similar to the "Options in main UI - place into an accordion" option in the User interface menu
  4. Apply settings
  5. Reload the UI

Then, to actually use it:

  1. Go to the txt2img or img2img tabs
  2. There's an accordion there saying Postprocessing, click on it to reveal the selected postprocessing options, click on it again to hide them

Additional information

No response

w-e-w commented 9 months ago

interesting idea

by the way if you just want to do face restoration then https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/12912#discussioncomment-6880457

narukaze132 commented 9 months ago

@w-e-w I actually want to do ESRGAN upscaling, which is why this is an issue.

w-e-w commented 9 months ago

@narukaze132 if you just want to do upscale automatically after txt2img then that is a trick you can use Hires. fix but set Denoising strength to 0 and select the upscaler and scale you want it will perform the upscale but will skip the second pass

this method would only work in txt2img as Hires. fix is only available in txt2img