Gemini Omni 1.1 Flash Arrives in Gemini API With 4K Output, 40-Second Scene Extension, and a 60%-Faster Draft Tier
Google announced Gemini Omni 1.1 Flash on August 27, positioning the update as the point where generative video crosses from prototype to production for professional workflows. The model is available via the Gemini API and Google AI Studio.
The original Gemini Omni brought real-world reasoning to generative video creation. The 1.1 Flash update is narrower in scope and more practical: four specific capabilities that address the friction points developers have hit when trying to ship generative video in production environments.
Scene Extension Now Has Memory
The most significant change is scene extension context length. The original Omni referenced only the final second of prior video when extending a scene. Omni 1.1 Flash analyzes up to 10 seconds of context — a 10x increase.
In practice this closes a consistency gap that made longer generated sequences visually jarring. Lighting, subject continuity, and narrative coherence across extensions depend on the model having enough context to make consistent decisions. The new limit supports extensions of up to 10 seconds per increment with a cumulative maximum of 40 seconds — enough for a complete short clip or a meaningful scene beat. You chain increments; each one draws on the full preceding 10 seconds.
Keyframe Interpolation
First and last frame specification is now supported: give the model a starting frame and a target ending frame, and it generates the continuous video path between them. The use cases Google highlights are camera orbits, zoom transitions, and seamless loops.
This is the capability that turns generative video from a text-input novelty into something that integrates with existing production pipelines. Compositors and motion designers working with reference frames can treat the model as a generative in-betweener rather than a standalone prompt-to-clip box. It’s not a replacement for traditional animation tooling, but it closes a meaningful gap for storyboard iteration and transition generation.
Two-Tier Resolution Pricing
Omni 1.1 Flash introduces a 360p draft mode alongside the existing 720p standard. Google cites up to 60% faster generation and one-third the cost compared to 720p.
The economic logic is straightforward: the bottleneck in generative video workflows is iteration, not final render quality. A team that can prototype a spot in 360p, approve the direction, and then render the final cut in 1080p or 4K is operating a fundamentally different workflow than one paying full price on every speculative pass. The 360p tier is priced to make iteration cheap enough that creative approval happens before significant compute is spent.
On the output end, Omni 1.1 now supports 4K upscaling alongside 1080p, covering broadcast and streaming production specifications.
Competitive Context
Gemini Omni 1.1 Flash entered the Image-to-Video and Text-to-Video Arena leaderboards on August 27 alongside the API launch, on the same day Alibaba’s Wan 3.0 was added to the Video Edit Arena. Comparative scores on Arena are not yet available, but the additions put both models in direct head-to-head evaluation.
The competitive framing is clear: Wan 3.0 competes primarily on generation speed, with Alibaba claiming 30-second generation times. Omni 1.1 Flash is competing on controllability and production tooling. Scene extension with temporal context, keyframe-to-keyframe interpolation, and tiered resolution pricing answer questions about workflow integration rather than raw generation throughput. Whether that positioning holds once Arena head-to-head results come in will determine how seriously developers weight controllability versus speed at the current capability tier.