stabilityai

/

stable-video-diffusion-img2vid-xt-1-1

like 521

Hi,

What about creating a model with reversed videos? The idea is to propose an AI that generates a video from an image that ends the video instead of being the first frame. It should be quite easy to do:

1. Reverse the videos from the original dataset
2. Train the model the same way
3. Create a dedicated space that inverts the order of the generated images

Regards