Load your image and hit Queue
COG models will be downloaded on the first run
Use T5XXL models, fp8 or fp16. Same as SD3 or FLUX
When it starts processing, stop, adjust the crop and run again, don't change the resolution, only (crop position and offset)
You will find a note with 20 different type of camera motion, you can use it as a base and edit them
Output videos will be saved here ".\output\CogVideo"
For reference, it takes 460 seconds for me to complete the run using a 4090, for a 6-second, 24fps, 2160x1440 output video.
Description
FAQ
Comments (7)
Why didn't you let the nodes breathe but made them tight af? xD
Hello, ComfyUI couldn’t find
the workflowUpscale + Frame Interpolation
workflowCOGVIDEO nodes.
I searched on GitHub and Google with no results.
Any help would be appreciated. Thanks!
SAME
@randomchatter1234776 @Uberkoly Hi, AFAIK custom group nodes are currently not supported in the comfyUI v1 desktop app, this is working properly on a fresh windows+ComfyUI portable. I exported the workflow again with the current ComfyUI, You can retry to download it and see if its working for you
Download is missing a .json workflow.
You can drop the png into comfyUI just like json
Could you please add start frame > end frame to the workflow? Only part of the pipeline I'm having a hard time finding. This workflow works great!