search

Klace / stable-diffusion-webui-instruct-pix2pix

Extension for webui to run instruct-pix2pix
Other
320 stars 20 forks source link

Add as sub-tab in img2img tab? #1

Open Jonseed opened 1 year ago

Jonseed commented 1 year ago

Since this works very similarly to img2img and inpainting, maybe it makes sense to add it as a sub-tab on the img2img tab, like inpainting is. It could just be another sub-tab called "instruct pix2pix" or "ip2p" for short. That way you don't have to duplicate all the functionality of that tab in your extension in a whole separate tab, and the user could easily copy the image between sub-tabs.

Klace commented 1 year ago

I was working on that and got stuck using the processing.py pipeline. Maybe in the future!

Jonseed commented 1 year ago

Does it not work with the processing.py pipeline having it in the img2img tab?

Klace commented 1 year ago

Does it not work with the processing.py pipeline having it in the img2img tab?

It's not that simple unfortunately, the denoiser and conditioning for the model is different than in txt2img or img2img. This is probably why nobody implemented it quickly and is the reason my extension and previous fork is pretty raw.

I'm still working on this so maybe one day we can merge it into the core code and have it not be an extension. Then we could get more functionality without reinventing the wheel, but if it's just me trying to figure it out may take awhile 😂

I will work on both adding new features to the extension and trying to get it into the processing pipeline, but progress will be slow unless there's a breakthrough with the processing.