SD-XL "4 output setup" help

#79
by ParaDim - opened

Can someone explain why SD-XL doesn't output 4 images with each generation as standard like Stable diffusion 2.1 and earlier models did? All the spaces on here have number of "iterations" but it seems like this has replaced "inference steps" as setting 25 iterations still only generates one image per prompt? How can I set-up SD-XL 1.0 to generate 4 images with every prompt I input?

Thanks!

I've not tried it but the pipeline call has this parameter...
num_images_per_prompt (int, optional, defaults to 1) β€” The number of images to generate per prompt.

images = base_pipe(prompt = prompt,
                 ... 
                 num_images_per_prompt=4)

I hope you've got the VRAM for it :-)

ParaDim changed discussion status to closed

Sign up or log in to comment