Gemini Omni 1.1 Flash adds keyframe control and 40s scene extension
Google's video model now takes first and last frames plus 3-second refs; 360p drafts cost a third of 720p, so iterate cheap and upscale to 4K.
Copy markdown
Scene extension now runs to 40 seconds
The model reads up to 10 seconds of prior footage — a jump from one second — and extends clips in 10-second steps to a 40-second cumulative length, holding motion and story across each cut. Long shots without stitching them together in post.
First and last frame control
Give it a starting and ending frame and it generates the video in between — deliberate camera moves, clean transitions, and seamless loops you can actually storyboard. Keyframes turn a slot-machine prompt into predictable output.
360p drafts at a third the cost
Draft at 360p for ~$0.034/second — roughly a third of 720p's ~$0.101 and up to 60% faster — then upscale the keepers to 1080p (~$0.152) or 4K (~$0.304). Burn cheap iterations, pay for polish only on winners.
Reference video locks character consistency
Feed up to three seconds of reference video into the multimodal input and it carries visual context and character identity into new shots — the missing piece for keeping one person or style coherent across a sequence.
Live now via API
Model ID gemini-omni-1.1-flash is available today in Google AI Studio and the Gemini Enterprise Agent Platform API, plus Google Flow and the Gemini app for AI Plus/Pro/Ultra subscribers. Input runs $1.50 per 1M tokens; video output $17.50 per 1M.