Google launches Gemini Omni 1.2 Flash for video

Google has launched Gemini Omni 1.2 Flash, a video generation model that creates clips of up to 15 seconds, lets you edit them through conversation and costs 30% less than the previous version. It is now generally available through the Gemini API and Gemini for businesses.

The announced price is $0.07 per second at 720p resolution. The idea is that you can generate a quick draft, fix only the details that need changing and save 1080p or 4K conversion for the versions you actually want to publish.

Longer videos that are easier to correct

Omni 1.2 Flash lets you choose durations between 3 and 15 seconds, in one-second increments. Version 1.1 supported up to 10 seconds.

That makes it possible to create a brief product demonstration, a scene with a beginning and an ending or an ad for social media in a single generation. It is not long enough to replace a complete video, but it can reduce the need to stitch together many short clips.

You can also continue editing the same clip with successive instructions. For example:

• Change the background of a scene.
• Replace a product or modify the lighting.
• Change the weather or the video format.
• Create variations for different audiences and platforms.

Instead of generating everything again from scratch, the model tries to preserve the performance, framing and main elements of the original video.

More control over characters, products and brands

The model accepts images of characters, products or styles, as well as motion clips of up to three seconds and audio references. According to Google, Omni 1.2 Flash preserves these details more effectively throughout the video.

In practice, this could help turn product photos into catalog videos, create ads with a specific visual identity or generate clothing demonstrations. Google shows a 15-second virtual try-on as an example. It combines a model's photo with front and back images of three outfits, preserving the model's identity and the details of each garment during 360-degree spins.

It also improves consistency between frames. That means fewer unintended changes to faces, text, packaging, colors, objects and settings when the camera moves or the scene gets longer.

More natural dialogue and camera movements

Gemini Omni 1.2 Flash includes improvements to speech, lip-syncing and the handling of multiple characters. Google says it can assign different voices to each character and maintain more expressive synchronization in dialogue scenes.

These features may appear in tools such as Flow, Gemini for businesses and Google Vids, where they are used to create explainers, avatar videos, narrations and ads.

The model also aims to follow complex camera movements, such as first-person flights, orbits and camera pullbacks, with fewer unexpected cuts and abrupt changes in scale. Google still acknowledges ongoing challenges in areas such as fast-paced sports, frequent script changes and complex physics.

What changes for you

For a creator or a business, the main difference is not just that the video lasts five seconds longer. The process becomes more iterative: you generate one version, correct a specific part and keep the rest.

That can reduce the cost of producing many variations of an ad or adapting the same content to different formats. For developers, the API makes it possible to integrate this workflow into applications and campaigns at scale, with Google Cloud enterprise infrastructure and an invisible SynthID watermark in the results.

Developers can try it in Google AI Studio and use the Gemini API. Businesses have access through Google Cloud. In Google's applications, availability depends on the product: Google Vids lets you create 15-second clips at no additional cost with a Google or Google Workspace account, while the feature in the Gemini app is being rolled out to Google AI Ultra subscribers.

The point to watch will be whether these consistency improvements hold up in more complex scenes and during rapid changes. For now, Gemini Omni 1.2 Flash targets a very specific use case: producing more video variations, with less repetition and at a lower cost.