stabilityai

/

stable-diffusion-3-medium

like 3.57k

same for me

MBP m2 32G
work fine with Preview (Nightly)
https://pytorch.org/get-started/locally/

I also have this issue. Although I believe the issue should lie with comfyuis latest update, since after updating, not even my sd1.5 or sdxl models will produce images. (They turn out completely black while sd3 produces multicolored noise)

Making a new conda environment for a Comfy UI with Nightly installed pytorch solved the issue for me. Thanks @alexfang888 !

Same issue on M3. Indeed installing new venv from scratch did the trick!
I used this script: https://github.com/comfyanonymous/ComfyUI/discussions/3283
You can skip the git clones if you have an existing installation.

Diffusers version of the model works fine using Diffusers on an M3

Getting strange images on Mac using ComfyUI. Using the same settings on colab, I get normal results.
Tried different versions of PyTorch, nothing helped. Any ideas?

T5 fp16, cfg 4, 40 steps