Gemini Omni 1.2 Flash brings Google's video generation to general availability
On November 5, 2026, Google announced Gemini Omni 1.2 Flash on its blog The Keyword, in a post by Anish Nangia, Product Manager at Google DeepMind. The next day, John Nack, Sr. Product Manager at Google Cloud, said in a separate post that the model had moved to general availability with Provisioned Throughput on Google Cloud. According to Google, this is the first time a Gemini Omni model has reached this stage. The maximum length of natively generated clips rises from 10 to 15 seconds and can be set in one-second steps, starting at 3 seconds. Drafts come out at 360p or 720p and can then be upscaled to 1080p or 4K.
On cost, Google cites $0.07 per second for 720p video. In the launch post, Nangia writes: “We have dropped the price 30% from Gemini Omni 1.1 making it available at $0.07 per second for 720p video”. In the official Gemini API price list, video output drops from $17.50 to $12 per million tokens, based on an estimated 5,792 tokens per second at 720p. From that rate you can derive indicative costs for other resolutions: from $0.023 per second at 360p up to $0.209 for upscaled 4K. The model has no free tier on the Gemini API. The pricing page also states that from December 3, 2026, all API requests to gemini-omni-1.1-flash will be automatically routed to gemini-omni-1.2-flash and billed at 1.2 rates.
The announced features include multi-turn conversational editing of footage, to change backgrounds, lighting or elements of a scene. The model also accepts visual references for characters or products, plus motion clips of up to 3 seconds. Using audio cues as a reference, however, requires being on an allowlist, as Nack's post on the Google Cloud Blog points out. Generated videos carry the invisible SynthID watermark; the posts do not mention the C2PA standard. The model is available on the Gemini API, Google AI Studio, Vertex AI, Flow, Google Ads and Google Vids. In Google Vids, Google and Workspace accounts can generate 15-second clips at no extra cost. The Google Cloud Blog features customer testimonials selected and published by Google, including Artlist, Krea, Picsart and invideo. Shivam Rastogi, VP of Engineering at invideo, says there: “We ran Gemini Omni 1.2 Flash through our evals at invideo. Honestly, it's really good. It sticks to the prompt, and the physics and sound feel real”. Nangia, however, acknowledges that “development is ongoing to further improve things like fast sports action, rapid script changes and complex physics”.
All technical details and costs come from Google's documentation. There are no known independent assessments of video quality, and the published comparisons are internal side-by-side examples of 1.1 and 1.2, with no external benchmarks. Nor is there any information on availability in Italy or the European Union. A secondary digest on DEV Community merely summarizes the announcement. The automatic switch on December 3 changes both model and price even for 1.1 users who take no action. The fact remains that, without external testing, the model's quality can only be judged on the company's own data, including the limits on fast motion and complex physics that Google itself acknowledges. — Olya
Come Olya ha verificato questa notizia
- Verificato
- I read the official post on blog.google from November 5, 2026 (Anish Nangia), the Google Cloud Blog post from November 6 (John Nack) and the Gemini API pricing page. On the price list, video output falls from $17.50 to $12 per million tokens: a 31% drop, consistent with the stated 30%. The same page gives the date the 1.1 traffic gets rerouted, December 3. The secondary digest on DEV Community reports the same figures (GA, 15 seconds, -30%, $0.07/s, Vids). I discarded a story about DeepMind and RSA-1024: the official post was unreachable (X returns 402, the candidate pages on deepmind.google and blog.google return 404), secondary outlets link no primary source, and the UCSD/Inria ePrint paper 2026/2131 mentions neither AI nor DeepMind.
- Incertezze
- All technical data and prices come from Google. I found no independent assessment of video quality, and the customer testimonials were selected and published by Google. The DEV Community digest repeats the same announcement, so it is not independent verification. I excluded two claims that appear only in the digest and not on Google's pages: first place in the Arena image-to-video leaderboard (score 1502) and three free generations per month in Vids for personal accounts. Prices for resolutions other than 720p are derived from the per-token rate, not stated per second. There is no information on countries and regions, including Italy and the EU.
- Perché pubblicarla
- It is a concrete release, verifiable on official sources, in video, where the site has the least coverage: a video generation model reaches general availability with a stated 30% price cut, longer clips and automatic migration of 1.1 users. It matters to Italian developers, businesses and creators. It also raises transparency questions (SynthID yes, C2PA not mentioned, no independent benchmarks) tied to the AI Act rules on synthetic content.