Gemini Omni extends video generation and editing to fifteen seconds
Google has released Gemini Omni 1.2 Flash for video generation and conversational editing. Clips can run up to fifteen seconds, with users revising the same scene rather than starting again. The company describes combined visual, motion and audio references, while acknowledging remaining difficulties with fast action and complex physics. Access varies across its developer and consumer products.
Artificial Intelligence··Night
Fifteen-second scenes stay editable
Google has made Gemini Omni 1.2 Flash, its video generation and editing model, generally available through its developer interface and business services. The release extends the native clip limit from ten to fifteen seconds. Users can select durations from three to fifteen seconds in one-second steps. Conversational editing keeps work on the same clip: a later instruction can change the product, background, lighting, weather or aspect ratio. Google says this also works across the full fifteen-second duration. The change lets a sequence of revisions remain attached to an existing scene instead of requiring a separate fresh generation for each adjustment.[1], [2]
Images, motion and sound guide generation
Inputs can combine images describing characters, products or style with audio cues and motion references lasting up to three seconds. Google’s examples include assigning voices to different speakers, matching lip movements and maintaining camera motion. Users can supply the first and last frames to generate the movement between them over three to ten seconds. They can also extend a clip by three to ten seconds in a turn, to a combined maximum of thirty seconds. Google identifies fast sports action, rapid changes to a script and complex physical interactions as continuing development areas. These remain documented limitations of the released video workflow.[1], [2]
Availability differs between Google products
The workflow allows lower-resolution drafts before selected results are upscaled. Generated output carries Google’s imperceptible SynthID watermark, and enterprise customers have a provisioned-throughput option for reserved processing capacity. Google lists AI Studio and the application programming interface for developers, Google Cloud for businesses and Flow for directing scenes and requesting conversational changes. It also describes fifteen-second generation in Google Vids. Access inside the Gemini application is rolling out to Google AI Ultra subscribers. From December three, requests addressed to the earlier Omni one point one Flash version will automatically route to the new version. That transition brings existing applications onto the updated model without requiring them to change their requested model name.[1], [2]
