Google Launches Gemini Omni 1.2 Flash With 15-Second AI Video

Longer clips and better references

Gemini Omni 1.1 was capped at 10 seconds. Version 1.2 Flash lets users pick any length from 3 to 15 seconds in one-second steps. According to Google, this makes it easier to produce product promos, social media ads and short story formats in a single pass.

The model also sticks more closely to reference material. Besides character, product or style images, users can supply motion clips of up to three seconds and audio references. Google says faces, clothing, product packaging, fonts and brand colors should stay consistent across camera angles. One example is a virtual try-on video built from a model photo plus front and back images of garments.

Editing without starting over

Multi-turn editing lets users change a finished video with follow-up prompts instead of regenerating it. Backgrounds, lighting, weather, the product shown or the aspect ratio can all be adjusted this way. Google also says character, object and scene consistency has improved, with fewer shape changes and lost details in fast motion or camera transitions.

Dialogue sounds more natural, lip sync is more accurate and the model separates the voices of several speakers better. Smooth camera moves such as FPV flights, wide pull-backs and orbital shots should come with fewer unwanted cuts.

Availability

Besides the Gemini API and Google Cloud, the model is available in Google AI Studio, Flow, Google Vids and Google Ads. In the Gemini app, 15-second video generation is rolling out to Google AI Ultra subscribers.

All output carries Google's invisible SynthID watermark. Google acknowledges the model is still being improved for fast sports action, abrupt scene changes and footage with complex physics.