AUTOMATIC1111
img2img api ValueError: Coordinate 'right' is less than 'left'
Is there an existing issue for this?
- [X] I have searched the existing issues and checked the recent builds/commits
What happened?
if i use mask and set inpaint_full_res =true
pyton will tip ValueError: Coordinate 'right' is less than 'left'
Steps to reproduce the problem
- Go to ....
- Press ....
- ...
What should have happened?
i hope it can working
Commit where the problem happens
SHA-1: 5ab7f213bec2f816f9c5644becb32eb72c8ffb89
What platforms do you use to access the UI ?
Windows
What browsers do you use to access the UI ?
Google Chrome
Command Line Arguments
--api --listen
List of extensions
no
Console logs
no
Additional information
No response
This happens when you're inpainting but there's no mask. I noticed this as well, some bug remove the mask randomly, not sure how to reproduce this.
Gen picture and mask, have similar exception
Picture and mask
both are 512*512
Draw on the image mate. :)
But how set, that it is an inpant, not just img2img?
Like this. Note the size should be for your mask, so maybe double if you want it higher resolution - but not for the entire image, when you use inpaint.
@andupotorac here we discuss an api, how to create curl request for inpanting draw?
@andupotorac here we discuss an api, how to create curl request for inpanting draw?
Oh, didn't notice that. Can't help with the API.
I had this error, many times.
Today I'm using inpaint upload. Based on the testimony above I replaced transparency in my mask file with white, and the error was gone, but it's attempting to inpaint the white area. Disjointedly.
The problem here is the webui expects the mask input to be a binary mask (RGB, only white and black), not an alpha mask. That being said some conversion should ideally happen automatically to correct this.
Edit: Actually looks like some upstream breaking change with Gradio broke this. Will have a fix shortly.
Fixed by https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/12588
Thank you! First, we wrote a crutch so that in .png the mask would be converted to black and white format without opacity. Now nothing works, although in version commit 68f336bd994bed5442ad95bad6b6ad5564a5409a (HEAD -> master, tag: v1.5.1, origin/master, origin/HEAD) everything worked fine.
Now, I understand that the .png needs to be set to opacity instead of a black and white image?
It's supposed to work for both. If it doesn't then something is going wrong. Last I tried both worked for me as of 1.6.0.
It's supposed to work for both. If it doesn't then something is going wrong. Last I tried both worked for me as of 1.6.0.
You can try processing this image with the provided mask (both 1179x1280)
Spoiler
The same pictures work fine on the version:
commit 68f336bd994bed5442ad95bad6b6ad5564a5409a (HEAD -> master, tag: v1.5.1, origin/master, origin/HEAD)
Merge: a3ddf464 50973ec7
Author: AUTOMATIC1111 <[email protected]>
Date: Thu Jul 27 09:02:22 2023 +0300
Merge branch 'release_candidate' ```
Your mask image is RGBA, and furthermore the alpha channel is not "full on" (i.e. a value of 255) across the whole image. The right-hand side is partially transparent. If your convert the image to RGB or fix the alpha channel it will work properly. So, this is not a bug.
The previous behavior was not correct because it was unable to handle alpha masks, but as a side affect it was more tolerable to mistakes like this in images themselves. The behavior is fixed now but it means you must make sure your masks are made properly.