Google launches Gemini Omni 1.2 Flash with 15-second video generation

Google has made Gemini Omni 1.2 Flash generally available, expanding its AI video model with native clips of up to 15 seconds, conversational editing, and lower pricing. The company says the model costs $0.07 per second for 720p video, a 30% reduction from Gemini Omni 1.1 Flash.

The new version supports videos from three to 15 seconds in one-second increments. That extends the previous 10-second limit and gives marketing teams, designers, and other content creators more room for short ads, product demonstrations, and social media clips without stitching together separate generations.

Google is rolling out the model through the Gemini API, Google AI Studio, Google Cloud Console, Flow, Google Vids, Google Ads, and the Gemini app for Google AI Ultra subscribers. Google says all generated videos receive an imperceptible SynthID watermark.

Editing without starting again

The central addition is multi-turn conversational editing. Users can generate a clip and then ask the model to change specific elements, such as a background, product, lighting condition, weather, or aspect ratio. Google says those edits can apply to a full 15-second clip while preserving the original action and framing.

That approach could reduce the need to regenerate an entire video after a small mistake or creative change. Google recommends a staged workflow: create early versions in 360p or 720p, revise individual details through follow-up prompts, and upscale the selected final version to 1080p or 4K.

The model also accepts several types of references. According to Google, users can provide character or product images, style images, audio cues, and motion clips up to three seconds long. The company says Omni 1.2 Flash better retains details such as faces, clothing, packaging text, and brand colors when camera angles or scenes change.

For commercial work, this is particularly relevant for videos that need to retain a recognizable product or person across several shots. Google presents examples involving virtual try-ons, product images, character sheets, and branded campaigns. These demonstrations are company examples rather than independent evaluations.

Focus on consistency, speech, and camera movement

Google also says the model reduces visual drift, in which people, products, or scenes gradually change during a generated clip. It claims stronger consistency during transitions, clip extensions, and first-to-last-frame generation.

Other improvements target dialogue and motion. Gemini Omni 1.2 Flash is designed to produce clearer lip-sync, assign different voices to multiple characters, and create more natural-sounding speech. Google says it can also handle continuous camera moves, including FPV-style flythroughs, orbiting shots, and rapid zoom-outs, with fewer abrupt cuts.

The company acknowledges that the system still needs improvement in areas including fast sports action, rapid changes in scripts, and complex physics. Those limitations remain important for users who need predictable results in action-heavy scenes or tightly choreographed productions.

Google positions the release as an enterprise-ready offering, with Provisioned Throughput available through its cloud infrastructure. The lower price and revision-focused workflow may make the model more practical for teams producing numerous variants of short video assets, though the company’s quality claims will need to be tested across a wider range of real-world projects.

Sources

• Gemini Omni 1.2 Flash – Google
• Google’s Gemini Omni 1.2 Flash stretches AI video clips to 15 seconds and cuts pricing by 30 percent – The Decoder