lightx2v

/

Hy1.5-Quantized-Models

like 38

Wrong model i believe.
See here:
https://huggingface.co/lightx2v/Hy1.5-Quantized-Models/discussions/2#6925c53a2b7e56f36e6eefea

Or make sure you download those with ComfyUI in filename here: https://huggingface.co/lightx2v/Hy1.5-Quantized-Models/tree/main

Wrong model i believe.
See here:
https://huggingface.co/lightx2v/Hy1.5-Quantized-Models/discussions/2#6925c53a2b7e56f36e6eefea

Or make sure you download those with ComfyUI in filename here: https://huggingface.co/lightx2v/Hy1.5-Quantized-Models/tree/main

Thank you, but I can confirm that I did not use the wrong model in this case. The hy1.5 video GGUF q4_K_M quantized model performs very poorly. I am not certain of the reason, but the q4_K_M quantized models of wan2.1 and wan2.2 deliver results that are significantly better.

https://huggingface.co/jayn7/HunyuanVideo-1.5_T2V_480p-GGUF/discussions/1#69257dee2074853d55db4ec3

GGUF q4_K_M is quite "compressed", but will give it a try here as well.
And note there is a difference between distilled models (that can use cfg 1), and the regular models, that you can not run with cfg 1 to get good results, but rather full steps and full cgi (as per recommended settings)

The guy that made the GGUF models has both a folder for distilled and regular models