Google released Gemini Omni 1.2 Flash on 5 November, a video model that makes clips of up to 15 seconds instead of 10. Google Vids now runs on it, and anyone with a Google account can make three of those clips a month there for free. For developers, the model costs 30% less than Omni 1.1: $0.07 per second of 720p video.
Google says reference photos now keep a product or face the same for the whole clip
Vids is Google’s video editor in the browser, part of Workspace next to Docs and Slides. You open it at vids.new, describe a scene and get a clip. Upload a photo of a product, a person or a place, and Google says it stays consistent from the first frame to the last.
The new model also brings clearer speech with better lip-sync and fewer unwanted jump cuts. A clip can be edited in conversation too: ask for another background, different lighting or a vertical format, and the model changes only that instead of starting over.
Three clips a month is enough to try the tool, not to fill a content calendar. Personal users who want more need a paid Google AI plan. For organisations, Google’s Vids help page lists a monthly pool of 200 seconds of AI video on Workspace Business Starter and 500 seconds on Business Standard and Plus.
Developers can draft at 360p and upscale only the best takes to 4K
In the Gemini API and on Google Cloud, Omni 1.2 Flash is generally available, so Google supports it for production use. Its pricing page puts Omni 1.1 at about $0.10 per second, and Google says the savings are similar at other resolutions. The cheapest way to work is the one Google suggests: draft at 360p or 720p, pick the takes you like and pay to upscale only those to 1080p or 4K.
- Google Vids: up to three 15-second clips a month at no cost, for any Google account.
- Gemini app: 15-second videos, rolling out to Google AI Ultra subscribers.
- Gemini API and Google Cloud: model ID gemini-omni-1.2-flash, $0.07 per second of 720p video.
- Also in Flow, Google’s tool for filmmakers, and in Google Ads.
- Every clip carries SynthID, an invisible watermark that detection tools can read.
Google says Omni 1.2 Flash still needs work on fast sports action and complex physics
Google itself says work continues on “fast sports action, rapid script changes and complex physics”. It published no benchmark results for Omni 1.2 Flash either. Its other claims rest on its own demo clips, such as a side by side in which Omni 1.1 follows a reference image less closely. The Google Cloud announcement adds quotes from customers including Picsart, Photoroom and invideo. Invideo says it ran the model through its own evaluations, but it published no method or scores.
Vids works in the Netherlands, but Google’s API rules out editing uploaded footage in Europe
Google names no country limits for the free Vids clips. Its help page says the Gemini features in Vids work wherever Google Workspace accounts are available, and that includes the Netherlands. Many of them still work only in English. Dutch is listed for AI avatars and voiceovers, but Google does not say whether video prompts in Dutch are supported.
Developers in Europe face a firmer limit. The Gemini API documentation, which still describes Omni 1.1, says editing or extending uploaded videos is not available to users in the European Economic Area, Switzerland and the UK. Editing clips that the model made itself does work. Google has not said in which Google Cloud regions Omni 1.2 Flash runs, so check that first if your data has to stay in the EU.
Realistic AI video needs a visible label in the EU since 2 August
For the organisations DataNorth works with, the free clips are a cheap way to learn whether AI video suits explainers, onboarding material or short social teasers. Run that test from a company Workspace account rather than from staff’s personal Google accounts. Google says its Business and Enterprise plans come with central admin controls, which personal accounts lack.
Before a clip goes public, settle two rules. Do not generate real, recognisable people without their consent. And label anything that looks real. Since 2 August 2026, Article 50 of the EU AI Act requires organisations that publish a deepfake to disclose that it is artificial. That covers AI video of people, places or events that viewers could mistake for genuine footage. SynthID does not do this for you, because nobody watching can see it. A short caption such as “made with AI” does.
For more information, visit the official announcement of Gemini Omni 1.2 Flash on the Google blog.
