[Bug]: Image loaded in Inpaint is out of reach + Loras, Embeddings, etc. appear dark in txt2img

hansolocambo commented 1 year ago

Is there an existing issue for this?

[X] I have searched the existing issues and checked the recent builds/commits

What happened?

I just installed your fork to test its UI and I found a few things that I'll list here. I've been using it for only a few minutes so I might find more. But the face that I cannot inpaint, when it's about 99% of the workflow, is kind of a game killer.

Issue 01 : One cannot browse Extensions (Loras, etc.) using the roll button. Only can one middle click and drag to slide the Loras, or use the scroll bar(s)

Issue 02 : Extensions appear normal in img2img but dark in txt2img

Issue 03 : Image loaded in img2img appears fine in the Image Source tab. But once loaded in inpaint, it can't be seen.

Setting the inpaint to full screen (great option) I can see the image, but it's cut in half and at the bottom of the screen. Using your Move button only slides the image left right... so basically inpaint doesn't work at all.

Last minute edit : Issue04 : the great ! extension canvas-zoom does not work with your UI. It enables the user, with well thought shortcuts, to zoom in and out or pan the image loaded in img2img or inpaint. Image then appears in the foreground of the everything in the UI (except the inpaint brush controls). You can then paint, change the brush size, etc : all with shortcuts. Super useful extension for heavy inpainters. I just realized that sadly this also doesn't work with your fork :/

Steps to reproduce the problem

use the middle roll of your mouse above Extensions. They don't scroll.
Go to txt2img, open the Loras, Emebeddings, etc. Tab. They appear normal in img2img, dark in txt2img
Drag and drop an image in inpaint. Or send it from txt2img or img2img. Image does not appear in inpaint. Only in full screen, but then the image is cut in half. Inpaint done in 2048x1152, which could be the issue ? (but it works fine with other Stable Diffusion I tried.

Great looking UI so far. Really good job you did there. I hope you'll find the few things that are still causing issues. Thanks for the hard work. I'll be sure to follow the development. Sincerely. Hansolocambo.

What should have happened?

scrolling as it should any part of the UI using the roll button.
extensions should never be dimmed for some reason.
inpaint is everything. Without a solid UI there I don't see any other option than back to Automatic1111 for the moment.

Cheers.

Commit where the problem happens

txt2img > Prompt extensions (Loras, hypernetworks, etc.)

What Python version are you running on ?

Python 3.10.6

What platforms do you use to access the UI ?

Windows 11 x64

What device are you running WebUI on?

Nvidia GPUs (RTX 3090)

What browsers do you use to access the UI ?

Mozilla Firefox / Chrome

Command Line Arguments

none. I just installed it and ran it once only so far, and without arguments.

List of extensions

canvas-zoom https://github.com/richrobber2/canvas-zoom.git ddetailer https://github.com/dustysys/ddetailer.git openpose-editor https://github.com/fkunn1326/openpose-editor.gi sd-webui-3d-open-pose-editor https://github.com/nonnonstop/sd-webui-3d-open-pose-editor.git sd-webui-aspect-ratio-helper https://github.com/thomasasfk/sd-webui-aspect-ratio-helper.git sd-webui-controlnet https://github.com/Mikubill/sd-webui-controlnet.git

Console logs

Please provide some info about how to provide you with the info you need. Most of us are users, non-tech savvy people ;) People might begin to write a bug report and abandon as they can't understand what's asked. 

If it's a .log file I can't seem to find any (Voidtools Everything) in my Stable Diffusion drive, except those created by Photoshop. Sorry but I can't find out how to get the log file.

Additional information

No response

P.S : I tried your UI in Chrome and it's even worse. There, the top left menu that replaces the tabs system in Automatic1111, is nowhere to be seen. I enlarge Chrome on more than one screen only to be sure. So on Chrome I can't even go to img2img tab to see if inpaint behaves any differently.

Good luck with the dev. Thanks again for this really good looking UI (b^-^)b

rushuna86 commented 1 year ago

the inpaint issue is canvas zoom. That extension fks up this whole fork's inpainting system, most of canvas zoom features are built into this fork so you don't really need it. And only fix is to do a clean install and never install Canvas Zoom. If you are going to reuse extensions and embedding folders from current install. Need to make sure you have also deleted the Canvas Zoom folder from under Extension Folder.

heroofthebeach commented 1 year ago

I just ran the inpainting issue as well. Deleting Canvas Zoom didn't fix the problem, as expected, but then I tried deleting /venv/ and letting the startup repair itself. And that fixed it! My inpainting works again!

This seems like it could be a common problem for the fork as people migrate from the primary version, wanting to bring their workspace with them. It's fortunate that the fix is relatively painless.

As far as the other issues, I also had the darkened extension cards, but it apparently went away on its own long before I ran into the inpainting bug. I'm guessing it's another imported extension issue, or perhaps futzing with the display settings knocked something bad into place. Either way, it wasn't any kind of permanent damage.

For what it's worth, my scroll wheel works just fine in the extension card tray. I haven't had any navigation issues with the fork.

rushuna86 commented 1 year ago

but yeah clean install and never install "Canvas Zoom" or delete /venv/ and also deleting "Canvas Zoom" from under Extension folder and running again so it creates a new venv. Add all your loras embedding + run webUI with just the built-in extension. And see if any of your problems still happen. And then add back your extensions slowly. That's how I originally found Canvas Zoom being incompat with this fork.

anapnoe / stable-diffusion-webui-ux