Text-to-Image
Diffusers
Safetensors
PixArtAlphaPipeline
Pixart-α
LCM

Same code for PixArt-XL-2-1024-MS works with PixArt-LCM-XL-2-1024-MS?

#2
by Softology - opened

I got the PixArt-XL-2-1024-MS working with your example code. When I change the model to use PixArt-LCM-XL-2-1024-MS I get high contrast results nowhere near as good as the non LCM model.

Should the same code work for both? Or are there some extra tricks needed for the LCM model?

Here is PixArt-XL-2-1024-MS
PixArt-XL-2-1024-MS.png

And here is PixArt-LCM-XL-2-1024-MS
PixArt-LCM-XL-2-1024-MS.png

Any ideas why the LCM model would give such a high contrast result?

Thanks.

Oh, there may be a mistake in the sample here. Can u try this out?
image = pipe(prompt, guidance_scale=0., num_inference_steps=4).images[0]

Thanks for the quick reply Lawrence. That fixed it :)
shrek eating pizza [PixArt-alpha] 72896006 2.png

Softology changed discussion status to closed

Sign up or log in to comment