Gemini Omni Flash 1.2

Overview

Gemini Omni Flash 1.2, which Google writes as Gemini Omni 1.2 Flash, is the third release of Google DeepMind's multimodal video model, announced on 5 November 2026. The shape of the model has not changed: a prompt, images, or short clips go in, and video with synchronised audio comes out, after which follow-up prompts keep working on the same take. What changed is how much of a scene fits in one generation and how little it costs to get there.

What's new in Gemini Omni Flash 1.2

A single generation now runs up to fifteen seconds, set in whole seconds from three, where 1.1 stopped at ten. Conversational editing covers the whole clip, so swapping a product, changing the weather, or asking for the vertical version are all follow-ups on an approved take. References are organised around people: six images per request, up to three of them for each of two characters. And Google lists 720p at about $0.07 a second, roughly 30% below the 1.1 rate, with 360p drafts cheaper again.

The quieter gains are in what holds still. Google describes steadier faces, garments, packaging typography, and colour through a shot, fewer involuntary cuts inside a continuous camera move, more natural dialogue with clearer lip sync, and conditioning for more than one speaker. Three things are named as unfinished: fast sports action, scripts that change direction quickly, and complex physics.

Fifteen seconds changes what you write

Ten seconds is a fragment, so a 1.1 prompt tended to describe one gesture and lean on extension for the rest. Fifteen seconds is a beat with a beginning and an end, which means the prompt can carry a small structure: the approach, the turn, the arrival. Extension still exists for scenes that run past that, in turns of three to ten seconds up to thirty in total, reading the final ten seconds of the clip as its context. The practical difference is how often you need it.

Gemini Omni Flash 1.2 and Morphic

Gemini Omni Flash 1.2 is not in the Morphic catalogue yet. Gemini Omni Flash 1.1 is, in the video model picker next to Veo 3.1, Seedance 2.5, Kling, and the rest, so the family's prompt shape and native audio are available to work with today, with the same brief able to run through several models and land side by side on the Canvas. Every example on this page was made on Morphic with the closest live model, not with 1.2, and Google has said requests to 1.1 on its own API route to 1.2 from 3 December 2026. For the prompt mechanics, read the Gemini Omni Flash 1.2 guide, copy a starting point from the Gemini Omni Flash 1.2 prompt library, see what changed in the 1.2 versus 1.1 comparison, or open the Gemini Omni Flash 1.2 AI video generator.