AUTOMATIC1111 / stable-diffusion-webui

Stable Diffusion web UI
GNU Affero General Public License v3.0
143.54k stars 27.02k forks source link

[Bug]: Generating an image, then reusing its seed will ignore any new image generated #6692

Open Manimap opened 1 year ago

Manimap commented 1 year ago

Is there an existing issue for this?

What happened?

When you generate any image in txt2img (I didn't test img2img or inpaint), if you want to reuse the seed of an already generated image in the webui, it will refuse to show the end result of a new image if the seed is the same, even if you change parameters (for example resolution). Instead it shows the first generated image with the seed. It also doesn't save the new image anywhere.

Steps to reproduce the problem

1- Generate any image, for example "chair" at 256x256 2- Then reuse the seed and change the resolution to 512x256 (you can change anything outside of the seed). 3- The resulting image is still the one generated before at step 1 instead of being the new one at the 512x256. If you check if any image was saved, nothing was saved anywhere.

What should have happened?

It should generate a new image with the new parameters, even if the seed is the same.

Commit where the problem happens

d930020044875fa780c95f7bc1881c7077d4fa59

What platforms do you use to access UI ?

Windows

What browsers do you use to access the UI ?

Brave

Command Line Arguments

--xformers --no-half-vae --deepdanbooru

Additional information, context and logs

No response

f-rank commented 1 year ago

Getting the same thing here. Thought I was somehow generating the same stuff, even after changing prompt. It actually saves it out in the proper directory for me, but does not load it and keeps the last image displayed.

Manimap commented 1 year ago

Tested it just in case since a lot has changed since last time. And still same problem even with latest commit 645f4e7ef8c9d59deea7091a22373b2da2b780f2 .