Google rolls out Gemini Omni 1.2 Flash with 15-second video and conversational editing

What happened?

Google announced Gemini Omni 1.2 Flash on November 5 and said it was rolling out across the Gemini API, Google Vids, Gemini for business, Flow, the Gemini app and Google Ads. The product headline is longer, more controllable video: native generation and editing up to 15 seconds, up from 10 seconds in Omni 1.1, plus multi-turn conversational edits and stronger reference following. Google says the 720p price is $0.07 per second, a 30% reduction from Omni 1.1.

The AI Studio documentation identifies the API model as gemini-omni-1.2-flash. It accepts text, image and video inputs, produces synchronized video and native audio, supports conversational editing through the Interactions API, accepts up to six reference images and offers video extension, first/last-frame interpolation and output-resolution control. The default resolution is 720p. A request can select a duration from three to 15 seconds; a duration above 15 seconds returns an invalid-argument error.

The rollout spans several surfaces, but the access promise is not identical on each one. Google says developers can prototype in AI Studio and use the API, while Google Vids offers 15-second clips to people with a Google account or Workspace subscription. The Gemini app rollout is described for Google AI Ultra subscribers, and Google Ads is framed as a campaign workflow. These statements establish product paths, not universal access for every country, account or plan.

The pricing page adds the caveats a launch summary can hide. Gemini Omni 1.2 Flash is generally available to developers on the paid tier of the Gemini API. Billing is based on output tokens, with Google’s examples putting a 10-second 720p clip at about $0.70 and a 15-second clip at about $1.04. Google lists separate effective rates for 1080p and 4K “with upsampling,” not native 4K generation. The announcement also says Google is still working on fast sports action, rapid script changes and complex physics.

Source: Google ↗ · Source: Google AI Studio ↗ · Source: Google AI Studio ↗ · Source: Google AI Studio ↗

Why does it matter?

The useful change is workflow, not simply duration. Multi-turn editing can let a team revise a clip without restarting from a blank prompt, while reference images and camera controls aim at continuity. That reduces the cost of an iteration cycle, but it does not remove the need to inspect text, identity, motion, physics or audio. Google’s examples are curated demonstrations and its performance language is a vendor claim. The model card and docs establish what the product is designed to do; they do not establish a universal success rate. Editors should treat each demo as evidence of a workflow claim, not a substitute for representative footage under their own workload and constraints for launch planning.

The 4K distinction matters for production planning. A catalog phrase such as “up to 4K output” can sound like a native high-resolution generator, while the pricing page explicitly describes 1080p and 4K as upsampling for Omni 1.2 Flash. That may still be useful for an ad or social workflow, but it changes compute cost, detail expectations and comparisons with systems that render native pixels. Teams should record the requested resolution and whether the output was upscaled before comparing models.

Availability is equally product-specific. A developer may need the paid API tier; Gemini app access is described for Google AI Ultra subscribers; Vids and Workspace access have their own account conditions; and the API documentation lists regional restrictions for editing uploaded videos. An early Arena snapshot can provide independent human-preference context, but its small sample is not a technical benchmark. The defensible news is a real rollout with practical editing and pricing changes, not proof that every account or workflow has the same access or quality.

Source: Google ↗ · Source: Google AI Studio ↗ · Source: Google AI Studio ↗ · Source: Google AI Studio ↗