1 paper · 1 filter
Anran Qi, Changjian Li, Adrien Bousseau +1
We address image-to-video generation with explicit user control over the final frame's disoccluded regions. Current image-to-video pipelines produce plausible motion but struggle t…