Hi! I was asked to create a lightweight workflow for low VRAM, so I made this one.
I hope it works well.
In this article I explain how to use it: https://civarchive.com/articles/27975
The models to dowload for it to work properly are these:
/models/text_encoders: https://huggingface.co/city96/umt5-xxl-encoder-gguf/blob/main/umt5-xxl-encoder-Q3_K_M.gguf
/model/unet:
SFW:
/models/vae: https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors
If you find it takes too long, lower the CFG to 1 and leave the negative prompts box blank.
Description
FAQ
Comments (7)
Worked perfectly, really fast. Thanks.
you're welcome!!
@saehara151 The only change I made was the updated model, from Mega v1 to Mega v12. Otherwise, really fast generation for short clips on demand.
@julianportillop599 I didn't know there was a new version, I'll download it, thanks for the info
@saehara151 and update the posting with a new link???
@ecgone634 And to say thank you for sharing a workflow for free that took you hours to build and you didn't have to do it without asking for money because it's time you could have used to do other things? It's not that hard to just Google the model name or ask any AI; seriously, it comes up first when you search for it on Google.
"Find Perfect Resolution" is a custom node
too bad theres a 90% chance of this thing drastically changing every face that looks under 40
I learned a ton getting this working, but I'm 90% sure the next step is to find something better bc this thing makes itself useless