Workflow broken in new ComfyUI version?
#15
by Bartholomheow - opened
After updating to comfy 0.24.1 the official workflow does not seem to be working consistently. Sometimes the generation is good, sometimes there's a lot of noise. Any idea as to why and how to fix?
Using Comfy Desktop on Windows
PyTorch: 2.10.0+cu130
CUDA: 13.0
z-anime-distill-8step-fp8, qwen3-4b-fp8, and the provided video autoencoder.
832x1216 size.
8 steps
cfg 1.0
euler
beta
No upscaling
Thanks