Gemini Omni 1.1 Flash adds 10-second video context
Google says Gemini Omni 1.1 Flash can reference up to 10 seconds of preceding video when extending a scene, rather than only the final second. It also adds keyframe interpolation and object editing for more controlled video generation.
TL;DR
- Scene extension now draws on up to 10 seconds of prior footage rather than Veo's final second, according to Google's scene-extension post.
- A generated shot can be bounded by a first and last frame, with continuous video interpolated between them, per Google's keyframe post.
- Natural-language instructions can swap objects or characters while retaining the surrounding scene, Google's editing post says.
- Final outputs can be upscaled to 1080p or 4K, according to Google's upscaling post.
- Google has made the release available through APIs, as Google's launch announcement describes.
The Gemini API pricing page puts 720p video output at roughly $0.10 per second. The API documentation tells clients to request URI delivery for generated files larger than 4 MB, avoiding response-payload limits.
Scene extension
Omni 1.1 conditions a continuation on up to 10 seconds of immediately preceding footage.
Each extension adds 10 seconds, and the launch page caps the accumulated sequence at 40 seconds. Its example continues a generation by passing previous_interaction_id; the API docs say each turn returns a new video while retaining prior video state.
Keyframes and natural-language edits
First-and-last-frame interpolation gives the model two fixed visual boundaries for a single generated shot.
Imported video is a separate control: Google's launch page allows up to three seconds of reference video in a multimodal input for visual context and character consistency. The documentation places one caveat on editing uploaded videos, which is unavailable in all regions.
360p drafts and 4K output
Google positions 360p as the iteration pass, claiming up to 60% faster generation and one-third the cost of its 720p mode. The same launch page offers 1080p and 4K output as the finishing pass.
Availability and model names
Google calls Omni 1.1 production-ready through the Gemini API in AI Studio, the Gemini Enterprise Agent Platform, and Google Flow for AI Plus, Pro, and Ultra subscribers. Scene extension is also rolling out in the Gemini app for those subscriber tiers.
The documentation still labels Gemini Omni Flash as preview and uses gemini-omni-flash-preview in its examples, while the launch post's extension sample uses gemini-omni-1.1-flash. Pika says the model is also available through its API Club in its API Club post.