This repo is for testing. Added CogVideo Dualtextencode. I have noticed it adds a lot more movement to the image to video with CogVideoXFun. It was just merged into Kijai main branch so try it out there instead https://github.com/kijai/ComfyUI-CogVideoXWrapper
Example here: kijai#61
Findings: It seems to add movement even when using only using a single t5 text encoder, but you can also add dual clip loader to load in t5 with clip_l and control the strength.
CogVideoX_Fun_00016.mp4
Here was the juggernaut x init image from the workflow:
Workflow to test it out here: