A lot of people ask me for my workflow, so here it is!
this workflow is for LOW VRAM
I advise you to use a lower fusionX GGUF model (Q4 or Q3) for a Vram of less than 10GB.
in "POSITIVE Selector" put 1 for Florence (auto prompt) + positive and 2 for positive only
To increase movement and avoid slow motion I recommend increasing the CGF to 3 or 4 but this reduces quality so increase the number of steps to 20.
Description
Wan GGUF + autoprompt + lora + upscaler
FAQ
Comments (6)
Thanks for the amazing workflow.
I was having trouble animating because of my AMD RX 6750 xt, and after a lot of testing I managed to make it work. It's very simple, but since I've never used ComfyUI before, it took me a while to learn.
fake not working on rtx 4050
VRAM issue maybe use FusionX GGUF Q3 and try lower resolution. I have a RTX3080 (10Go), my workflow is set to 832 maybe try 720 or less if not working)
Ran out of memory when regular VAE decoding
I have a 3060 and I'm using a GGFU model Q4
It used the other VAE, but with the image upscaler my ram didn't survive
maybe try lower res in 'height" node like 720 or lower
i don,'t found this node:
Note _O
CR Text Input Switch
Text _O