What are the requirements to download?
The description says to download Text encoders, Clip vision, Transformer, and Vae from https://huggingface.co/Kijai/WanVideo_comfy/tree/main. However there are 30+ options to choose from and I cannot get a clear answer on what is what from the internet. The README seems to say that TinyVAE, SkyReels, WanVideoFun; are text encoders due to being first but honestly I have no idea. I don't really want to put the wrong files in the wrong places and they all appear to be .safetensors files.
The installation says:
Text encoders to ComfyUI/models/text_encoders
Clip vision to ComfyUI/models/clip_vision
Transformer to ComfyUI/models/diffusion_models
Vae to ComfyUI/models/vae
The readme has the following files together:
1: TinyVAE, SkyReels, WanVideoFun;
2: Lightx2v, CausVid 14B, CFG and Step distill 14B;
3: CausVid 1.3B, AccVideo, Phantom, ATI, MiniMaxRemover, MAGREF;
4: CausVid LoRAs v1, v1.5, v2;
I am using comfyui with 8 gigs of vram, and 32g of RAM, I have 600g out of 2 terabytes on my storage. I was told that some downloads depend on your system. Any advice on which files to download would be appreciated.
I followed Benji's youtube video more or less, used his workflow, and the list of files was more or less:
magref into diffusion models (it's in the big list of files)
text_encoders folder
https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/text_encoders/umt5_xxl_fp16.safetensors
vae Wan2_1_VAE_bf16.safetensors into vae folder (in the big list)
https://huggingface.co/Kijai/WanVideo_comfy/blob/main/taew2_1.safetensors into vae_approx
into the vae_approx folder
Hope this helps. I used a resize from my FramePack workflow, because the images must be the same size. Just about to try it!
Thanks, its working now!