Model update

Gemini Omni 1.2 Flash:
15-second clips and
conversational edits

Google released the Gemini Omni 1.2 Flash video model on November 5, 2026: native 15-second clips, conversational editing and 30% cheaper 720p. We looked at what it means for shot length, references and your budget.

6 November 2026 4 min read

By: Steve KadasAI film, directing, workflow

In short: Google released the Gemini Omni 1.2 Flash video model on November 5, 2026. According to the company, clips now run up to 15 seconds natively, a finished clip can be edited through conversation, and the 720p price is down 30%. These are the company's claims, and no independent quality test exists yet.

15 slongest native clip

0.07 USDper second at 720p

6image references per request

What changed from 1.1 Flash?

The headline change in Gemini Omni 1.2 Flash is length: you can set a clip anywhere from 3 to 15 seconds, in whole seconds, with 10 as the default. Version 1.1 topped out at 10 seconds. The Decoder reports that conversational editing works across the full 15 seconds. According to Google, you can swap the background, lighting, camera angle, aspect ratio or product without regenerating the clip. Unite.AI says anything you don't mention stays as it was.

Reference handling got broader too. A single request takes up to six image references, at most three per character, for up to two characters. Inputs can include character images, product photos, motion clips up to 3 seconds and audio cues. The company says faces, clothing, packaging text and brand colours hold across different camera angles. In Google DeepMind's own comparison, 1.2 beat 1.1 by 62 to 38 on visual consistency and 63 to 37 on multi-reference identity. That is the company's own test, not an independent one.

On audio, Google promises more natural dialogue, cleaner lip sync and better separation of multiple speakers. The Decoder reports that the model handles continuous camera moves (FPV drone, zoom-out) with fewer jump cuts. The model is available in the Gemini API, AI Studio, Google Cloud Console, Vids, Google Ads and the Gemini app on Ultra. The Flow and Gemini app rollout is still in progress. It is generally available on Google Cloud, and outputs carry a SynthID watermark. Anyone can generate up to three 15-second videos a month for free in Google Vids.

What does a 15-second clip actually cost?

According to Google, Gemini Omni 1.2 Flash costs 0.07 USD per second at 720p, and The Next Web reports that 1.1 cost 0.10 USD per second in June. The cut is per second, so a longer clip still costs more in total. A developer blog (DEV Community) puts a 15-second 720p clip at ~1.04 USD, against ~1.01 USD for a 10-second clip on 1.1. By our own math, the raw output of a 30-second film at 720p comes to ~2.1 USD (2 × 15 s × 0.07 USD), before any discarded takes.

What does it mean for film planning?

A 15-second clip is not a 15-second shot. Our guide to shot length recommends 3 to 5 seconds per shot, so 15 seconds is really a scene with movement in it. A 30-second ad can now come together from 2 or 3 longer clips instead of 6 to 10 short ones, but the pacing will feel different. Native 15-second generation is also not the same as video extension. Extension is a separate feature that adds 3 to 10 seconds per turn, up to 30 seconds in total.

What's new	What it means for planning
Native 3–15 second clips	Plan a scene with a start and end state, not a single shot.
6 image references, max 3 per character	Allocate the six slots on your shot list up front: with 3 images each for two characters, no slot is left for a product photo. The basics: character consistency.
Voice references on an allowlist	Only build on your own voice once you actually have access.
360p/720p draft, 1080p/4K final	Three tiers: draft, conversational fixes on the same clip, and only the winner gets upscaled.
Continuous camera moves	Describe a long move with distance and time: camera moves in meters and seconds.

Draft at 360p or 720p, and only upscale the clip you pick to 1080p or 4K.
Ask for fixes in conversation, then check that what you didn't mention really stayed the same.
Budget in seconds: clip length translates directly into cost.
Test the pacing on the three free 15-second videos a month in Vids before you pay through the API.
Check consistency with your own references, since 62 to 38 is the company's own number.

According to its homepage, Weblock AI Studio currently writes prompts for Kling 3.0, Seedance 2.5, Veo 3.1, Wan 3.0 and Nano Banana Pro, and Gemini Omni 1.2 Flash is not on that list. Generation runs in your own account, and the Studio shows a credit estimate before every paid step.

What don't we know yet about Gemini Omni 1.2 Flash?

Google hasn't published per-second prices for Gemini Omni 1.2 Flash at 1080p or 4K, only mentioning "similar savings." The approval criteria for voice references are unknown, and so is the full scope of the Flow and Gemini app rollout. There is no independent quality test yet, and no comparison with Veo 3.1.

Sources

Related in Learn

How long should a shot be? Pacing for 15, 30 and 60 s A shot runs roughly three to five seconds, and the shots add up to exactly the film's length. H… Character consistency: so it isn't a different face in every shot The most common failure in AI video: by the second scene, your lead is somebody else. The fix i… Camera moves in metres and seconds, not adjectives Ask a video model for a slow push-in and you often get a frozen shot. Here is how to write a ca…