Google has released Gemini Omni 1.2 Flash, a video-generation and editing model that extends native clip creation to 15 seconds, adds multi-turn conversational editing and lowers pricing by 30% compared with its predecessor. The company says the model is generally available through the Gemini API, Google Cloud and Gemini for business.
The update is aimed at a practical constraint that has shaped much of the first generation of AI-video tooling: short clips are useful, but they often leave too little time to tell a complete story or make a usable product demonstration. Gemini Omni 1.2 Flash can generate and edit videos from three to 15 seconds, with duration specified in one-second increments. That is an increase from the 10-second maximum offered by Omni 1.1.
Longer clips, with an emphasis on iteration
Fifteen seconds is still a compact format, but it is long enough for a short social advertisement, a product explainer or a self-contained visual sequence. More important than the added five seconds may be the workflow Google is positioning around it. Rather than requiring a user to discard a nearly successful result and produce an entirely new clip, the model supports follow-up edits in a conversation.
Google says that approach lets users make adjustments without restarting from the original prompt. In video production, small inconsistencies can be costly: a product may change appearance, a character may lose continuity, or a requested motion may not appear as intended. The company is framing conversational edits as a way to address those problems in subsequent turns while retaining the broader direction of the clip.
The release also emphasizes stronger reference adherence. Users can guide a generation with combinations of character, product or style images, short motion clips of up to three seconds and audio cues, according to Google. The stated goal is to make it easier to carry specific visual direction into a result rather than relying only on text instructions.
Lower prices change the production calculation
Google has set the listed price for 720p video at $0.07 per second, describing that as a 30% reduction from Gemini Omni 1.1. Price is not just a purchasing detail for generative video. It affects how freely creative and production teams can experiment, how many iterations they can afford, and whether a tool is used for occasional concepting or becomes part of a repeatable workflow.
The company suggests a staged process: draft quickly at 360p or 720p, use follow-up editing to fix details, then upscale selected creations to 1080p or 4K. That sequence reflects a familiar production logic. Lower-resolution versions can be used to test composition and motion before computationally heavier finishing work is applied to the strongest candidate.
For developers and businesses, the combination of lower unit pricing and a model available through an API matters as much as the headline feature. It makes the system easier to incorporate into products, internal content pipelines and automated creative workflows. Google also says the release is available in Google Vids and Gemini for business, putting the capabilities into tools intended for workplace use as well as developer platforms.
What the launch signals
Gemini Omni 1.2 Flash is not presented as a single breakthrough in visual realism. Its significance is more operational: longer native clips, controls that can be revised through conversation, more inputs for directing output and a lower cost per second. Together, those changes target the gap between a striking generated sample and a tool that can be used repeatedly by people working to a deadline.
That distinction is increasingly important as AI-video systems move from demonstrations toward production. A marketing team, educator or software developer may value reliable adjustment and recognizable product references more than an isolated cinematic result. The ability to specify a duration, give the model multiple kinds of reference material and continue editing a clip makes the tool more structured than a one-shot prompt interface.
There are still limits inherent in the format. A 15-second cap constrains the kinds of stories that can be told in one generation, and Google’s announcement does not eliminate the need for human review of outputs. But for work that is already designed around short-form video, the added duration may reduce the need to assemble several separate AI-generated fragments.
Google’s pricing and availability choices suggest that it sees fast, lower-cost generation as central to adoption. By putting the model in its API, cloud offering and business-facing products at launch, the company is pursuing both builders who want to embed video generation and organizations that want to use it directly. The result is a model update focused less on novelty than on making short AI-generated video more editable, more directed and more economical at scale.
This article is based on reporting by Google AI Blog. Read the original article.
Originally published on blog.google
