Google ships Gemini Omni Flash to general availability as gemini-omni-1.1-flash

Google has moved its fast, conversational video generation and editing model out of preview. On August 27, 2026, the Gemini API changelog announced the general-availability release of gemini-omni-1.1-flash, closing out a preview period that began earlier this year.

What's new

The GA release ships with several capabilities that were not available in the preview endpoint:

• Video extension — the model can continue an existing clip past its current end point, either through a dedicated extend task or directly from a text prompt.
• Interpolation between two images — using the image_to_video task with up to two images, the model generates a video that transitions from a first frame to a last frame, effectively animating a start-and-end pair.
• Resolution control — a new resolution parameter inside video_config supports 360p, 720p (the default), 1080p, and 4K output, with the two higher tiers produced via upscaling rather than native generation.

Google's changelog describes the release as "the GA version of our fast, conversational video generation and editing model." Alongside the GA announcement, Google set a hard deprecation date of September 30, 2026 for the old preview endpoint, giving developers a five-week window to migrate before it stops working.

Context

Omni Flash sits inside Google's broader Gemini video push, which has moved quickly from research demos toward production-grade, API-addressable tooling. The GA release lands in the same week Google made other developer-facing changes to the Gemini line, including new usage-limit mechanics for Gemini Notebook — a sign that the company is actively iterating on both the consumer and developer sides of its Gemini surface as competition in AI video intensifies.

Unlike a one-shot text-to-video generator, Omni Flash's positioning around "conversational" editing — extend, interpolate, adjust resolution — targets iterative workflows where a creator refines a clip over several turns rather than generating once and starting over. That's a different product bet than pure generation-quality plays from labs like Runway or Luma, leaning instead on speed and editability.

Why it matters

General availability is a meaningful signal in itself: preview endpoints carry no uptime or deprecation guarantees, while GA status typically comes with production SLAs, which matters to any team building a shipped product on top of the model rather than a demo. Forcing preview users off the old endpoint by September 30 also shows Google is confident enough in the GA model's stability and pricing to stop supporting the interim version, rather than running both in parallel indefinitely.

More broadly, the release adds another credible, API-accessible entrant to the fast-moving AI video generation and editing category, alongside OpenAI's Sora work and dedicated video specialists like Runway, Luma, and Pika. For developers building agentic or conversational video tools, having extension and interpolation as first-class, resolution-configurable API tasks — rather than capabilities bolted on after the fact — lowers the engineering lift needed to ship iterative video editing features.