Workflow Wan Vace Workflow.

The thing with Minimax H3 is that it regenerates the entire background and environment that's surrounding the character based on your prompt. So you have to give it really explicit prompt details otherwise it's not the same environment like wan vace is.
 
the workflow works perfectly.

But it does use a lot of RAM.

I used a rtx pro 6000 with 175gb RAM for a 40 sec video and i kept getting OOM errors. the VRAM isnt the issue but the RAM.
Any fixes? like optimizing the workflow for lower vram usage.

and if possible a Vast.Ai tutorial aswell, just found that the prices for renting are much cheaper and i can get larger ram capacities aswell.

i tried the same instructions of the runpod version and tried on vast.ai but i couldnt get jupyter to open.
 
This workflow is amazing. Thank you
Can I ask you one question?
I tried many times but it seems like lower body isn’t turning out as prompt says. For example wearing white pants or towel. Even I write prompt completely nude. Can you help me?
 
Which workflow are you using? For nudity you need to add nsfw loras. Did you add that?
 
wan22_vace_inpaint_int8_convrot_sam3.json I used this workflow
And I used Lora
NSFW-22-L-e8
genitals_helper_v1.0_e219
 
you still need to describe what the masked area replaced with. e.g "woman dancing fully nude revealing her breasts and vagina"

and on negative prompt add specifically what you want removed. e.g "black shirt, gray pants,"
 
i am running runpod and it works but now i dont know how to add nsfw loras... how do i add them and where can i find them
 
this is awesome was waiting for someone to come and post it again
does anyone have the ltx workflow?
 
what's the requirements for running this workflow, spec wise? I know that you would need an nvidia GPU
 
You must log in or register to reply here.
Back
Top