Gemini Omni 1.2 Flash adds 15-second video and lower API output rates

Google’s November 5 release of Gemini Omni 1.2 Flash extends video generation and editing to 15 seconds and lowers video-output rates. For game teams, it could supply promotional concepts or storyboard previews. The Gemini Developer API offers the model on its paid tier; generated shots still need review against the game they will represent.

Fifteen seconds, with different limits for each mode

The model specifications list generation and editing from 3 to 15 seconds. First-and-last-frame interpolation remains at 3 to 10 seconds. Supplying two endpoint images does not unlock a 15-second transition.

Reference images can guide characters, props and style. Conversational editing can then change a background or lighting in an existing clip. The Developer API guide currently excludes uploaded audio inputs and audio references, even though the model can generate synchronized sound. Editing or extending uploaded videos is also unavailable in the EEA, Switzerland and the UK; edits to model-generated clips remain supported there.

Google says the update improves reference following and continuity. Its ramen-shop comparison also acknowledges changes to small details such as cats and utensils. For a character reveal or environment preview, review appearance, lettering and shot continuity before using the result.

Budget for video output and inputs separately

The Developer API price table lists $12 per million video-output tokens. At 720p, that works out to approximately $0.07 per second, or $1.04 for 15 seconds of video output. Inputs and text output add charges, and this model has no free API tier.

Drafting at 360p or 720p and upscaling selected shots to 1080p or 4K is another option, with different output costs. Keep generated concept footage distinct from actual gameplay when planning a campaign through the marketing and launch category.