Google Releases Gemini Omni 1.2 Flash Video Model for General Availability
Google released Gemini Omni 1.2 Flash on November 5, its first generally available (GA) video generation model, producing up to 15-second clips with audio at $0.07 per second for 720p—roughly 30% cheaper than v1.1, or about $1.04 per 15-second clip. The update adds multi-turn conversational editing of the same video and consistency control from up to six reference images plus a 3-second motion clip, and replaces Veo on the Gemini API.
Related Entities
Full Report
宝玉 (@dotey) · 2026-11-06 21
Google Releases Gemini Omni 1.2 Flash: Up to 15 Seconds per Video, About $1 per 720p Clip, Available to Developers and Enterprises Today
On November 5, Google released the video model Gemini Omni 1.2 Flash. It can generate videos up to 15 seconds long with audio in a single pass, at 30% lower cost than the previous 1.1 version, priced at $0.07 per second for 720p.
This is also the first time Omni has reached general availability (GA), meaning it can be used directly in production environments. Omni is the model Google announced at its I/O conference in May this year, and Google compares it to Nano Banana: like editing an image, you refine a video round by round through chat.
[How much does a 15-second video cost]
Based on the official pricing page, a 15-second 720p clip costs about $1.04. Version 1.1 was about $0.10 per second with a maximum of 10 seconds, so previously $1 bought 10 seconds; now it buys 15 seconds.
Draft in 360p first for about $0.35 per 15 seconds; once satisfied, render in high definition: about $1.56 for 1080p and about $3.13 for 4K. Durations range from 3 to 15 seconds, selected in whole seconds.
The API free tier is not available; a paid account is required. Starting December 3, requests calling 1.1 will be automatically routed to 1.2 and billed at 1.2 pricing.
[Who can use it and where]
Developers can try it in Google AI Studio today and integrate it via the Gemini API; enterprises can access it through Google Cloud.
For general users: in Google Vids (Google's video creation tool), anyone with a Google account can generate 15-second clips at no extra charge; in the Gemini App it is gradually rolling out to AI Ultra subscribers.
It is also available in the AI filmmaking tool Flow; the blog post does not specify which subscription tier is required, though version 1.1 was open to AI Plus, Pro, and Ultra users. In Google Ads, it can be used to produce advertising videos in bulk.
[How it improves on 1.1]
First, length goes from 10 seconds to 15 seconds, so a complete product introduction or social media ad can be generated in one pass.
Second, you can make successive requests against the same video: change the product, change the background, adjust the lighting, switch between landscape and portrait, without starting over. Several versions of the same ad can be produced for different audiences.
Third, it follows reference materials more faithfully. You can provide up to 6 reference images (people, products, styles) plus an action video of up to 3 seconds, and Google says faces, clothing, text on packaging, and brand colors stay consistent even when the camera angle changes.
In Google's own demos, one model photo plus front and back images of three outfits produced a 15-second turning try-on video; another changed backgrounds 5 times in a row while the character's performance and composition stayed the same. Lip sync for dialogue is also more accurate.
[Compared with Veo, Sora, and others]
On the Gemini API, Omni has already replaced Veo: the Veo 3.1 preview was retired on October 22, and the official guidance is for developers to migrate to Omni 1.2 or use the GA version of Veo on Google Cloud.
On OpenAI's side, Sora is no longer an option. According to OpenAI's help center, the Sora App shut down on April 26 and the API was discontinued on September 24.
On the Runway developer platform, which sells models from multiple vendors, per-second prices for video with audio are: Veo 3.1 $0.40, Veo 3.1 Fast $0.15, ByteDance Seedance 2.0 (720p) $0.30, and Kuaishou Kling 4 (720p) about $0.16 at list price. Omni 1.1 sells there for $0.10, the same as Google's own price. Compared with these, Omni 1.2's $0.07 is the lowest.
For now, claims about visual quality come only from Google and partners such as Runway and Picsart, and the demos were selected by Google itself. Google says fast sports action and complex physics still need improvement. The official documentation states that only English has been fully evaluated, with no guarantee of results for Chinese prompts; uploading your own audio as a reference is not yet supported in the Gemini API.
Official blog: https://t.co/F94nWAaZo0 API pricing: https://t.co/AfcmTu3YD9
