Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

3x Controlnet not working using Batch images #1625

Closed
Trikos opened this issue Aug 31, 2024 · 8 comments
Closed

3x Controlnet not working using Batch images #1625

Trikos opened this issue Aug 31, 2024 · 8 comments

Comments

@Trikos
Copy link

Trikos commented Aug 31, 2024

Hi, Im using the forge to create img2vid to create animations because ti is super fast instead of using a1111.
The steps I take are:
img2img - Batch tab, select my path, enable all 3 controller and select Batch tab using the same path as before. In my folder there are 5 different images, when I start generating it only takes the first image and iterate 5 times on that ignoring all the others.
What can I do to fix it?
Im using Euler a, 20 step, 512x910, CFG 3, denoising 1, clipskip 2, ckpt aniverse 1.5
Controlnet: pixel perfect, control net is more important.

Some commands to enable when I start the webui maybe? Or some settings?

Thank you, amazing job

@westNeighbor
Copy link

The controlnet batch never works since I first started to use Forge a year ago. I mostly want it on the txt2img.

@DenOfEquity
Copy link
Collaborator

Trikos, I might be misunderstanding your process, but it seems you are setting batch within img2img and then in controlnet as well? The img2img batch will send the images to all the controlnets as needed. No need to set up the batch within controlnet too. Doing it this way is working for me.

westNeighbor, for batching within controlnet you have to set Batch size to match the number of images sent to controlnet via Batch Folder tab. Batch count doesn't work for this. And Batch Upload tab in controlnet seems to be broken at the moment.

@westNeighbor
Copy link

It looks my bad. It does work by set Batch size. The logic is different with A111 which is set by Batch count.

@westNeighbor
Copy link

@DenOfEquity Is there a way to fix the seed number for all batch controlnet images? Currently it will increase the seed by 1 whether I fix the seed or not?

@westNeighbor
Copy link

westNeighbor commented Aug 31, 2024

Trikos, I might be misunderstanding your process, but it seems you are setting batch within img2img and then in controlnet as well? The img2img batch will send the images to all the controlnets as needed. No need to set up the batch within controlnet too. Doing it this way is working for me.

westNeighbor, for batching within controlnet you have to set Batch size to match the number of images sent to controlnet via Batch Folder tab. Batch count doesn't work for this. And Batch Upload tab in controlnet seems to be broken at the moment.

It accumulates all the batch controlnet images into the vram, so at most ~30 batch images for a 16GB video card. If I have 150 batch images, it takes more than 5 hours for my 4080 card. I think for the controlnet batch, the official A1111 way is the way to go. Should start an issue for this problem.

@DenOfEquity
Copy link
Collaborator

I don't know any way to fix the seed for this. There is a setting that is named like it should control it (controlnet_increment_seed_during_batch), but it's not referenced anywhere in the code. And looking at old Forge, not used there either.

@westNeighbor
Copy link

I don't know any way to fix the seed for this. There is a setting that is named like it should control it (controlnet_increment_seed_during_batch), but it's not referenced anywhere in the code. And looking at old Forge, not used there either.

Screenshot 2024-08-28 130031
The new A1111 add the function, see above image. @DenOfEquity

@Trikos
Copy link
Author

Trikos commented Sep 12, 2024

Ok it worked, I guess there was a glitch or something else cause It always used the same images. I did a clean install.
Thank you all

@Trikos Trikos closed this as completed Sep 12, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants