VINCIE-NExT: Unlocking Video Editing from Images via In-Context Modeling.

arXiv:2610.12104v1 Announce Type: cross Abstract: Building a capable video editor remains significantly harder than a video generator: editing requires (source, instruction, edited) triplets that are prohibitively expensive to annotate and difficult to synthes