Google has released Gemini Omni 1.2 Flash, a new version of its fast video-generation model that makes clips up to 15 seconds long and costs 30% less than the version it replaces. The model is now generally available through the Gemini API, Google Cloud and Gemini for business, the company announced on 5 November 2026.
Gemini Omni is Google’s model for generating and editing video from text, images and other references. The 1.1 version capped clips at 10 seconds. The new version stretches that to 15 seconds — and lets creators pick any length from 3 to 15 seconds in one-second steps, so a clip can be timed to fit a social ad slot or a product demo rather than trimmed afterwards.
The price cut is the other half of the story. Gemini Omni 1.2 Flash costs $0.07 per second for 720p video, down 30% from Omni 1.1. That works out at about $0.10 a second for the older model, by simple arithmetic. At the new rate, a full 15-second clip at 720p costs $1.05. Google said similar savings apply at other resolutions, and businesses running large volumes can buy Provisioned Throughput on Google Cloud.
Editing without starting over
The longer clips arrive with a second change that matters as much for day-to-day use: multi-turn conversational editing. Instead of re-generating a whole clip to fix one detail, a creator can keep the same clip and ask for changes in plain language — swap a product, change the background, adjust the lighting or switch the aspect ratio — turn by turn. Google says the editing works across full 15-second clips, which makes it practical to spin out ad variants for different audiences without re-briefing the model each time.
Reference handling has also improved. The model follows image, video and audio references more closely, holding faces, garments, packaging text and brand colours steady across camera angles. Google’s example: one model photo combined with front-and-back shots of three outfits produces a 15-second virtual try-on sequence, with the model’s identity and each garment’s cut and drape staying consistent through 360-degree turns. Audio references are gated behind allowlisting, so not every developer can use them yet.
Speech gets better too. Google says the model produces more fluid dialogue, clearer lip-sync and distinct voices for multiple speakers — the kind of detail that shows up in multi-character scenes in Flow, video explainers in Gemini for business, and avatar-led videos in Google Vids.
Where it is available
Developers can use the model through the Gemini API, prototyping in Google AI Studio, and businesses can build with it on Google Cloud. It is the first time Gemini Omni has been generally available on Google Cloud, backed by Google’s enterprise infrastructure and security.
For people who do not write code, the model is rolling out across Google’s own products: 15-second clips in Google Vids at no extra cost to Google account and Workspace users, scene direction and multi-turn edits in Flow, 15-second videos in Google Ads, and a gradual rollout in the Gemini app, reaching Google AI Ultra subscribers first.
Every output carries an imperceptible SynthID watermark, Google’s marker for AI-generated content.
Why it matters beyond the US
Fifteen seconds is the length of a standard social ad, and a 30% price cut changes the arithmetic for the businesses that make video in volume. A small retailer in Karachi, Lagos or Manila can now turn existing product photos into short video ads, try-on clips or catalogue spins without a shoot — the same workflow Google’s cloud blog highlights for retailers and marketplaces. Platforms including Artlist, Krea, Picsart, Photoroom and Runway are already building the model into their products, which puts it in front of creators who never touch an API.
The timing fits a busy week for AI video. OpenAI’s ChatGPT Images 2.6 added animated GIFs days earlier, and Tencent’s Hy4 model arrived with computer-use agents — a sign that the race is now about what these models can do in real workflows, not just what they can render.
What it still cannot do
Google is open about the gaps. Fast sports action, rapid script changes and complex physics remain works in progress, the company said, and it published no quantitative benchmarks — only side-by-side examples showing the new model following references more closely than 1.1. Anyone planning a campaign around fast-moving action should test the model first rather than trust the demo clips.
Still, for the everyday videos businesses make in volume — product clips, try-ons and ads — 15 seconds at $0.07 a second is a concrete step: longer, cheaper, and editable in conversation. Google said it will keep improving the model based on user feedback.
