I've created this Wan Video Text2Video workflow using ComfyUI's native Wan nodes.
Key Features:
1st pass - generating vanilla video
2nd pass upscaling using user selected upscaler and frame interpolation to 32/60FPS
This workflow comes pre loaded with my Wan2.1 RunPod template:
https://civarchive.com/articles/11960/wan14b-in-1-click-with-workflows-included-runpod-template
Description
Added TeaCache support for the non native ComfyUI workflow - generation should be much faster!
Removed Kijai's clip encoder and added the native ComfyUI clip encoder (split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors)
download it here:
https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/tree/main/split_files/text_encoders
FAQ
Comments (9)
For anyone getting "TypeError: expected Tensor as element 1 in argument 0, but got NoneType"
There's a bug in the latest commit Kijai made,
Navigate to the WanVideoWrapper custom node folder (ComfyUI/custom_nodes/ComfyUI-WanVideoWrapper)
Run git checkout bd31044
Restart ComfyUI
I will remove this comment once it's fixed.
Thanks for the workflow! TeaCache speed up is crazy. 12s/it down to 4.5. I have been getting sort of blurry videos out of it though
What's your settings?
Work almost the same for me.
I guess it's just the upscaling part to be honest. I'm using the default settings from your workflow and it each of the upscaled outputs beyond the initial one are very blurry edges and lose tons of sharpness. Ended up just taking the initial output and running it through TopazAI and it seemed to retain a lot more clarity.
Same here. Amazing workflow but for me also every output after the initial one is extremely blurry and without any sharpness in the 60FPS workflow. Any ideas how to fix this?
DJDABass try dropping the cfg down to 1?
"mat1 and mat2 shapes cannot be multiplied"
Please make sure you're using the correct clip.
yeah I had this yesterday, it means one of the models is the wrong one.