Video extension, sometimes called video extend, is an AI video generation feature that adds new frames to the beginning or end of an existing video clip, continuing the visual content, motion, and scene coherently beyond the original clip's duration. Rather than generating a video from scratch, extension uses the existing footage as a conditioning anchor and generates additional moments that flow naturally from it, effectively lengthening a clip while maintaining consistency with the original material.

Video extension works by conditioning the generation process on the frames at the boundary of the existing clip, using the established visual state, motion trajectories, lighting, and scene composition as constraints on what the newly generated frames should contain. Extending from the end of a clip continues the motion and scene forward in time; extending from the beginning generates footage that plausibly preceded the existing clip. The quality of the extension depends on how coherently the model can infer the continuation or antecedent of the visual content, and on how well the boundary frames provide sufficient information to anchor the generated material. Extension is a practical workflow tool for situations where a generated clip is almost the right length but needs a few additional seconds, or where a specific clip needs to be extended to fill a fixed duration in an edit. Repeated extension can compound small inconsistencies, so very long extensions from a short original clip tend to drift more than single-step extensions.

Compose on Morphic includes video extension as part of its timeline-based generation workflow, allowing creators to lengthen clips by generating additional frames that continue the visual content naturally. This is particularly useful when a generated clip cuts off before a motion or scene has fully resolved, enabling the creator to extend through to a more satisfying conclusion without regenerating the entire clip from scratch.