What changes compared with the previous version
Google has announced the general availability of Gemini Omni 1.2 Flash, a model dedicated to video generation and editing. According to the source, the new version expands functionality while maintaining a lower generation cost than in the past, with a direct impact on those who produce video content through automated tools.
The most concrete figure concerns duration: whereas Gemini Omni 1.1 stopped at 10 seconds, version 1.2 Flash allows native clips of 3 to 15 seconds to be specified, in increments of one second at a time. The stated price is $0.07 per second of 720p video, about 30% less than the previous model. These are the two parameters on which the difference from the previous generation hinges, because together they affect the breadth of content that can be produced and the cost per second produced.
Distribution channels and editing functions
Distribution is structured across several fronts. Developers and businesses can access the model via the Gemini API and Gemini for business. On the consumer side, availability comes through Google Vids, the Gemini app reserved for Google AI Ultra subscribers, and Flow. The simultaneous presence across professional tools and mass-market consumer products indicates that Google aims to cover both productive and individual creative use, although the source does not specify timing or activation methods for each channel.
In terms of features, the model accepts visual and audio references: images of characters, products, and styles, short video clips, and audio can be combined to maintain consistency of appearance—for example, of clothing or typography—even when the camera angle changes. The source also reports a reduction in visual artifacts between frames and shots, with the aim of making continuous camera movements smoother. For editing, multi-turn modification on a single clip is provided, with natural-language instructions to adjust background, lighting, and format. On the audio side, the model handles lip-sync for multiple speakers and the generation of natural dialogue.
Content tracking and unresolved issues
Google specifies that the generated content incorporates SynthID, a watermark not visible to the naked eye. This is relevant because it concerns the identification of synthetic videos, but the source does not clarify how the watermark is detected by end users or what tools are made available for verification.
Several points remain to be examined. The material does not indicate the contractual terms and usage limits for Google AI Ultra subscribers, nor the commercial terms for companies that will use the Gemini API. It is not specified whether model 1.2 Flash entirely replaces version 1.1 or whether the two coexist, and no geographic restrictions are specified. This is information useful for assessing the actual scope of the launch, which is currently not covered by the source.
