Update — August 28, 2026: fal's H3 Max landed a day later and places Gemini Omni Flash second on fal's internal Elo table (2060 vs 2080) at 26.70s average latency against H3 Max's 3.49s.
TL;DR: On August 27, 2026, Google shipped Gemini Omni 1.1 Flash — an update to the "anything in, anything out" world model behind Gemini Omni Flash. The headline changes: a 10-second video context window that lets extensions chain up to 40 seconds total, first-and-last-frame controls for storyboarding a shot precisely, video style matching, and a draft-then-upscale pricing model — $0.03 per second at 360p, with optional upscaling up to 4K. Google says it tops the Text-to-Video Arena at 1,515 points. It's live now in Google AI Studio, Google Flow, and the Gemini app.
| Question | Answer |
|---|---|
| What is it? | Version update to Gemini Omni Flash — same "anything in, anything out" model family, new controls |
| Announced when? | August 27, 2026, via Logan Kilpatrick (@OfficialLoganK), Google, and Google AI Studio on X |
| What's new? | 10s video context, 40s max extensions, first/last-frame control, video style matching, video references |
| Draft price | $0.03 per second of video at 360p |
| Upscale ceiling | Up to 4K, as a separate optional step |
| Benchmark claim | Text-to-Video Arena, 1,515 points (Google's own claim, not independently verified) |
| Where to use it | Google AI Studio, Gemini API, Google Flow, Gemini app |
What Changed From Gemini Omni Flash to 1.1
Gemini Omni Flash launched June 30, 2026 as Google's cost-efficient model for video generation and conversational editing, positioned as a developer-first tool through AI Studio and the Gemini API. Two months later, Gemini Omni 1.1 Flash keeps that same positioning and pricing philosophy but adds four concrete capabilities that change what you can actually build with it.
Logan Kilpatrick, who leads AI Studio at Google, framed the update directly on X:
"this is an update to our anything in, anything out world model. It now supports: 360p drafts, 4K upsamplers, up to 10 seconds of video context when extending, video extensions in 10 second increments, video references."
Google's own framing on the announcement account was similar, describing 1.1 as delivering "a new suite of creative capabilities and controls for developers" — specifically calling out the ability to extend scenes, specify the starting and ending frames of a shot, and add video references.
Here's what each of those means in practice:
- 10-second video context for extensions. The original Omni Flash generated clips in a single pass. 1.1 lets you extend an existing clip by feeding the model up to 10 seconds of prior context, so the continuation stays visually and motion-consistent with what came before — rather than restarting from a fresh prompt that happens to loosely match.
- Up to 40 seconds total via chained extensions. Four extension passes off an initial generation gets you to the 40-second ceiling. This is still short of a full scene by film standards, but it's a meaningful jump from the single-shot clips most text-to-video APIs (including the original Omni Flash) ship with by default.
- First-and-last-frame controls. You can now specify exactly how a shot starts and ends, and let the model fill in the motion between those two anchor frames. This is a practical storyboarding tool — instead of prompting blind and hoping the ending lands where you want it, you pin both ends and generate the transition.
- Video style matching and video references. You can hand the model a reference clip and ask it to match visual style — lighting, color grade, camera movement — across a new generation, or maintain a consistent character or setting across separate cuts using video references.
The Draft-Then-Upscale Workflow: Why $0.03/s Matters
The pricing structure is the part most worth internalizing if you're building on this. Gemini Omni 1.1 Flash generates draft videos at 360p for $0.03 per second, with upscaling to higher resolutions — up to 4K — available as a separate step.
That two-tier structure is a deliberate cost-optimization pattern, not just a technical limitation:
- Iterate cheaply at 360p. At $0.03/s, a 10-second draft costs $0.30. You can run a dozen prompt variations, test first/last-frame combinations, and try different style references for a few dollars total.
- Lock a cut before you pay for resolution. Only send the clip you actually intend to ship through the 4K upscaler. This mirrors the proxy-edit workflow from traditional video production — cut on low-res proxies, conform to full resolution only for the final export.
- Extensions inherit the same draft-first logic. Since each 10-second extension is itself a generation call, chaining toward the 40-second ceiling is cheaper to explore at 360p before committing to a final upscaled render.
For teams building automated video pipelines — product demos, marketing assets, social content — this pricing shape means the real cost driver isn't the base rate, it's how disciplined your team is about upscaling only what survives review. A workflow that upscales every draft defeats the purpose of the tier.
The Text-to-Video Arena Claim
Google says Gemini Omni 1.1 Flash tops the Text-to-Video Arena leaderboard at 1,515 points. Developers cited in Google's announcement reportedly praised the model's consistency for cohesive storytelling — a claim that lines up with the first/last-frame and video-reference additions, both of which are specifically aimed at keeping longer sequences visually coherent.
Worth being precise about what this claim is and isn't: it's Google's own reported result from its August 27, 2026 announcement. explainx.ai has not independently verified the score against the live Arena leaderboard, and leaderboard rankings shift as new models are submitted and re-scored. Treat the 1,515 figure as a vendor-reported benchmark claim, not a confirmed, stable ranking — the same caution worth applying to any single-day "we're #1" announcement from a model lab.
Where Gemini Omni 1.1 Flash Is Available
Google shipped 1.1 across the same three surfaces the original Omni Flash and its consumer spin-offs have used:
- Google AI Studio — the no-code playground at aistudio.google.com, where you can test extensions, first/last-frame prompts, and style matching without writing code.
- Gemini API — for production integration, using the same API patterns as the original Gemini Omni Flash, with the new context and frame-control parameters added to the video generation call.
- Google Flow — Google's creative production surface for AI-assisted filmmaking workflows.
- Gemini app — the consumer surface, the same one that powered the free 10-video Omni promo in late July and August, and Google Photos' Video Remix template edits.
If you're already integrated with the original Gemini Omni Flash via the API, the update is additive — existing calls should keep working, with the new context window, frame controls, and style-matching parameters available as opt-in additions rather than breaking changes to the base generation call.
Why This Matters for Builders
The original Gemini Omni Flash's pitch was workflow consolidation: video generation and conversational editing inside the same API-connected session developers already use for text and image tasks with Gemini. 1.1 doesn't change that pitch — it extends the ceiling on what a single generation session can produce.
Forty seconds via chained 10-second extensions is still well short of a full scene by traditional production standards, but it moves Omni out of "single establishing shot" territory and into "short sequence" territory — enough for a product demo beat, a social ad, or a multi-shot explainer clip. Combined with first/last-frame control, it's a genuinely different tool than the original single-pass generation: you can now storyboard a shot's boundaries first and let the model fill in the motion, which is a meaningfully more directable workflow than blind prompting.
The pricing model is the other half of the story. A $0.03/s draft tier with optional 4K upscaling gives teams a real lever for controlling video generation costs at scale — something that matters more as usage moves from experimentation to production pipelines. For a broader look at how Omni compares to the rest of the AI video landscape, see explainx.ai's complete guide to Sora, Runway, and Kling, and for the mechanics of how models like Gemini actually process video input, see Can Claude or LLMs Watch a Video?
Sources & References
- Logan Kilpatrick (@OfficialLoganK) on X — Gemini Omni 1.1 Flash announcement, August 27, 2026
- Google and Google AI Studio (@GoogleAIStudio) on X — Gemini Omni 1.1 Flash announcement, August 27, 2026
- Google AI Studio: aistudio.google.com
- Gemini API documentation: ai.google.dev
Related Reading
- Gemini Omni Flash: Google's AI Video Generation Model — the original June 30, 2026 launch this update builds on
- Gemini Omni video model spotted ahead of Google I/O 2026 — the earliest public sighting of the Omni family
- Gemini free 10 Omni videos promo — until Aug 4, 2026 — consumer Gemini app trial for non-subscribers
- Google Photos Video Remix — Gemini Omni consumer editing — template-driven Omni edits in Photos' Create tab
- Someone played an AI video inside Google Sheets, Calendar, and Chat using Antigravity — Omni's Interactions API pushed into unexpected surfaces
- Can Claude or LLMs Watch a Video? — how Gemini and other models actually process video input
- Video Generation AI: Sora, Runway, Kling — Complete Guide
Gemini Omni 1.1 Flash was announced by Google on August 27, 2026, via Logan Kilpatrick, Google, and Google AI Studio on X. Pricing, benchmark claims, and availability details reflect the announcement as of that date and may change as Google updates its documentation.