Very simple, easy to see, easy to use, the design is very human
UNET: https://huggingface.co/QuantStack/Wan2.2-I2V-A14B-GGUF/tree/main
BLINK LORAS: https://civarchive.com/search?q=Blink&rating=all
UPSCALE MODEL: https://huggingface.co/dtarnow/UPscaler/blob/main/RealESRGAN_x2plus.pth
Description
FAQ
Comments (15)
How did you fix the noise problem in generated videos?
Try increase the megapixel at the resolution calculate node. I set the default to 0.5 mp for speed. Also your start image should be high quality and high resolution. I also use a simple image upscale with RealESRGAN_x2plus and VFI frame interpolation for the final video
thank you!
how do i adjust length of video ?
Hi, i downloaded everything but i get a CUDA error. Are there parameters i could change to try to fix this?
i guess its comfyui issue, chk with chatgpt with ur error. if in runpod, better switch to compatible pod, if local, need to up/dw cuda. though i am not that expert, just learning
it could be a sage attention issue, try disable the sage attention nodes
Doesn't work. Keep getting miising node error for UNET Low even after installing Wan2.2-I2V-A14B-LowNoise-Q6_K.gguf from the link provided
the results are very blurry. any tips for steps/cfg settings?
Perfect thank you. I was having trouble finding a workflow that had multiple lora support. NSFW time!
I have to ask what is probably a stupid question. Do I need to add a LoRa of the star of my video? When I run this workflow as it is, I get a single 4 second video of the starting image (headshot) just moving around a bit. It seems obvious that the model needs more than this but I don't see any reference to more photos nor a LoRa.
Pardon me - I figured it out. Thanks for your work in putting this workflow together.