Could CogVideoX1.5-5B-I2V theoretically be adapted into a video-to-video model without training?

[yiboz2001]

Feature request / 功能建议

Could CogVideoX1.5-5B-I2V theoretically be adapted into a video-to-video model without training?

Motivation / 动机

I noticed that the current video-to-video frameworks are primarily built upon T2V models, e.g., CogVideoX-2b and CogVideoX-5b. Since CogVideoX1.5-5B-I2V is capable of generating videos at arbitrary resolutions, I am curious if it could theoretically be adapted into a video-to-video model without additional training.

I would greatly appreciate any suggestions or insights from the community.

Your contribution / 您的贡献

none

[zRzRzRzRzRzRzR]

The translation of the given text is: "We can refer to the community's approach to implement V2V, and the T2V model can be modified. But we haven't tried I2V."