VideoMaMa++ keeps video mattes consistent across chunks
Preserve and refine tokens let sliding-window inference carry mattes across chunk boundaries without extra training.
Stories per day · 30 days
Relatedautoregressive1computer-vision1diffusion1gradient-decoupling1long-horizon1sgf-plus1video-generation1video-matting1
2 stories, newest first · one card per story
Stories tagged #temporal-consistency
SGF+: Decoupling Gradient Flows for Autoregressive Video Generation
SGF+ decouples video context writing from denoising, enabling 24-hour generation from 5-second training rollouts.