Google AI Releases Gemini Omni 1.1 Flash: 40-Second Scene Extension, First/Last Frame Control, and 4K Upscaling

Google has shipped Gemini Omni 1.1 Flash as a production update to its multimodal video generation and editing model. The release focuses on three concrete capability areas rather than a broad architectural overhaul, and the changes address some of the more practical limitations users encountered with the previous version.

The most significant shift is in how scene extension works. Previously, the model used only the final frame of an existing clip as context when generating a continuation. With 1.1 Flash, it can now analyze up to 10 seconds of prior footage before extending the scene. That wider temporal window helps the model maintain consistent lighting, motion direction, and subject positioning across the join point - issues that often created jarring cuts in earlier outputs. The maximum extension length also reaches 40 seconds in this release.

Frame-level control has been added through first and last frame pinning. Users can now specify both the opening and closing frames of a generated clip, giving them a more direct way to choreograph camera movement and scene composition without relying entirely on text prompts. A related addition allows existing video clips to be passed in as reference material for character consistency, which is useful for productions that need a recurring subject to look the same across multiple generated segments.

The 4K upscaling feature completes the update, letting lower-resolution outputs be brought up to higher display standards in a post-generation step. Gemini Omni 1.1 Flash sits within Google's broader Gemini model family, which has been positioning native video understanding and generation as a core capability rather than a separate pipeline. These targeted additions suggest Google is iterating toward more precise editorial control, moving the tool closer to something that fits into professional or semi-professional video workflows.