Gemini Omni 1.2 Flash: 15-second AI video at a 30% lower price

On November 5, Google DeepMind released Gemini Omni 1.2 Flash, a new version of its model for generating and editing video. The model makes clips up to 15 seconds long (up from 10 in Omni 1.1), lets you refine them in follow-up messages and is generally available in the Gemini API and Google Cloud, while the price per second of 720p video dropped by 30%.

What’s new in Gemini Omni 1.2 Flash?

According to Google’s announcement, the main changes are:

1. clips up to 15 seconds, with length adjustable in one-second steps from 3 to 15 seconds,
2. conversational editing: instead of regenerating everything, in later turns we can swap a product or the background, change the lighting, the weather or the aspect ratio,
3. better reference following: the model accepts images of a character, product or style and short motion clips (up to 3 seconds), and faces, clothing, text on packaging and brand colors are meant to stay the same across different camera shots,
4. less image “drift” between the first and the last frame, also when interpolating between keyframes and extending a clip,
5. more natural dialogue, clearer lip-sync and multiple speakers in one scene,
6. complex camera moves with fewer unwanted cuts.

Interestingly, Google suggests a specific workflow: draft in 360p or 720p, refine in conversation, and finally upscale to 1080p or 4K. Every clip gets an invisible SynthID watermark.

How much does Gemini Omni 1.2 Flash cost?

According to the Gemini API pricing page, the model is only on the paid tier (it isn’t on the free tier) and is billed in video tokens, at $12 per million. Per second, that works out to:

Resolution	Price per second
360p	approx. $0.023
720p	approx. $0.07
1080p (upscaling)	approx. $0.104
4K (upscaling)	approx. $0.209

So a full 15-second clip in 720p costs about $1.05. Omni 1.1 Flash cost $17.50 per million tokens, or about $0.10 per second of 720p. From December 3, 2026, all requests to 1.1 will be redirected to 1.2 and billed at the new rates.

How to try Gemini Omni 1.2 Flash?

Google lists several places:

1. Google AI Studio and the Gemini API: the model ID is gemini-omni-1.2-flash (documentation),
2. Google Cloud: in the console (Vertex AI Studio) and through the API, including Provisioned Throughput,
3. Flow: for directing scenes and refining them over multiple turns,
4. Google Vids: 15-second clips at no extra cost with a Google account or a Workspace subscription,
5. the Gemini app: rolling out gradually to Google AI Ultra subscribers,
6. Google Ads.

In the API, a clip is 16:9 or 9:16 and lasts 10 seconds by default. We can add up to 6 reference images (up to 3 per character, at most 2 characters) and extend the video by 3 to 10 seconds per turn, up to 30 seconds in total.

Why is Omni 1.2 getting so much attention?

Google’s post made the front page of Hacker News, where it collected over 300 points and over 200 comments. The discussion is mixed, though: some people praise the character consistency, others ask what 15 seconds of video is actually good for (the answer: mostly B-roll and ads), and still others complain about the confusing rollout across Google’s apps.

Then there’s the price: video with sound and conversational editing for about a dollar per 15-second 720p clip. And it’s not Google’s only generative visual model in recent weeks, because at the end of October DeepMind announced Genie 4.

What to watch out for?

First, the limitations Google lists itself: fast sports action, abrupt scenario changes and complex physics still need work.

Second, language. According to the documentation, English is fully supported and other languages have not been tested (test dialogue in your language on a few short clips before you plan a whole project around it).

Third, the differences between the apps and the API. The announcement mentions audio cues as references, but according to the API documentation, audio upload is not supported yet.

Fourth, the lack of numbers. Google has published no benchmark results or comparisons with other video models, so let’s treat the promises of better consistency with some skepticism until we see independent tests. By the way: a few days earlier, OpenAI added animated GIFs in Images 2.6 to ChatGPT.

FAQ

What is Gemini Omni 1.2 Flash?

It’s a Google DeepMind model for generating and editing video with sound, released on November 5, 2026. It makes clips from 3 to 15 seconds from text, images and short reference videos, and lets you refine them in follow-up messages.

Is Gemini Omni 1.2 Flash free?

Not in the Gemini API: the model is only on the paid tier, starting at about $0.023 per second in 360p. You can make 15-second clips at no extra cost in Google Vids, with a Google account or a Workspace subscription.

How is Omni 1.2 different from Omni 1.1?

Omni 1.2 makes clips up to 15 seconds instead of 10, follows references better, lets you edit video in conversation and is 30% cheaper per second of 720p. From December 3, 2026, requests to 1.1 will go to 1.2.