Updates·November 6, 2026, 03:21

Google's AI video model now makes 15-second clips at a lower price

AI-generated from the sources listed, with automated source checks.

Google has made its Gemini Omni 1.2 Flash video model generally available. Clips can now run up to 15 seconds, and the price per second of video has been cut by 30 percent.

On November 5, 2026, Google made its AI video model Gemini Omni 1.2 Flash generally available. The model creates video clips from text descriptions, and it can now produce clips between 3 and 15 seconds long in one-second increments. The previous limit was 10 seconds.

Cheaper to use

According to The Decoder, it costs about $0.07 per second of 720p video when the model is used through the Gemini API. An API is a technical access point that companies and developers use to connect the model to their own products. The previous price was about $0.10 per second, so the cut is 30 percent.

The model supports four resolutions: 360p, 720p, 1080p and 4K. Google describes a workflow in which users make quick drafts in 360p or 720p and then upscale the best clips to 1080p or 4K.

Multi-round editing

According to The Decoder, users can refine a clip over several rounds in a conversation. That could mean changing the background, lighting or aspect ratio without regenerating the entire clip. Google itself says a single detail can be fixed with a follow-up edit instead of redoing the whole clip.

Google also cites improvements in how well the model follows references such as characters, products, typography and style. This also applies to consistency between frames, lip sync and camera movements. All videos carry the SynthID watermark.

In Google DeepMind's own comparisons, which The Next Web reports on, testers preferred version 1.2 over 1.1 in 62 percent of cases when it came to visual consistency.

Where can you use it?

The model is available in the Gemini API, Google Cloud, Google Vids (at no extra cost), Flow, Google Ads Asset Studio and Gemini for business. In the Gemini app, it is available to Google AI Ultra subscribers. Provisioned Throughput, which is reserved capacity, is also available.

Limitations

Google says it is still working on fast sports action, rapid script changes and complex physics simulations, The Next Web writes. The same outlet notes that ByteDance's Seedance 2.5 can make 30-second clips, compared with Google's maximum of 15. The Next Web also writes that Google is the only American company with a model among the top ten on a widely used AI video leaderboard. The other nine spots go to Chinese models. The name of the leaderboard has not been disclosed.