stable-diffusion-webui
stable-diffusion-webui copied to clipboard
img2img tab layout update
A number of smaller changes to the img2img tab to be more consistent with the txt2img tab:
- Moved related options to be next to each other. Editor related options are now all right below the editor. the "Editor hints" help text is hidden on a seperate tab rather than the slightly hacky expandable that was there before
- moved other options to be consistent with their positions on txt2img tab
Before/After comparison:
Before:

After:

Also while we're here, I question the usefulness of the "Push to img2img input mask" Button, doesn't a mask need to have transparency? And SD is quite literally incapable of generating an image with transparency?
I didn't want to just remove it without asking but I can't imagine ever using it...
I question the usefulness of the "Push to img2img input mask" Button, doesn't a mask need to have transparency
I thought Push to img2img input mask just pushes the base image and sets the mode into "mask". Of course there'll be no transparency, the idea is that you yourself will then add a mask with "sketch" tool or whatever it's called.
I'll test this but looks like it's fine just moves things around. Anyone else tested it yet?
@jansteffen I think @codedealer is right, you still need to draw a mask, some things need documenting better and renaming to be easier to understand
Push to img2img input mask just pushes the base image and sets the mode into "mask"
Ok I see that now, but even then, you're replacing two clicks with one and dedicating an entire button to it, even though the mask mode toggle is right next to the normal "Push to img2img input" button
I had issues when I was testing this just, the masking didn't seem to work, @jansteffen could you confirm it works for you? it might have been the way I'm symlinking the webui and main repo github folders to my actual working directory, I'll carry on testing now I switched back to master quickly to show off the current state of img2img to Emad on twitter 🚀
Seems to work fine for me, both in "keep only masked area" and "regenerate only masked area" modes
This is "regenerate only masked area", on the output image only the river changed as expected:

Sorry for the delay @jansteffen , I've been busy on a call so I merged all the pending pr's to one branch to make it quicker to test, your changes look good, onto the next!
No worries, you're doing amazing work managin the crazy amount of traffic on this repo!