Gemini Omni 1.1 Flash adds scene extension, keyframes, and 4K finishing to the API (2026)

Google DeepMind released Gemini Omni 1.1 Flash on August 27, 2026, a production-ready update to its generative video model. Developers call gemini-omni-1.1-flash on the Gemini API and Google AI Studio; Gemini Enterprise Agent Platform lists gemini-omni-1.1-flash-preview. The update adds scene extension, first-and-last-frame interpolation, 360p drafts, 1080p/4K finishing, and short video references. 1 2 3

What launched

Signal	Confirmed detail	Why it matters
Model IDs	API/AI Studio: `gemini-omni-1.1-flash` (paid, generally available). Agent Platform: `gemini-omni-1.1-flash-preview` (preview, Aug 27, 2026). Prior `gemini-omni-flash-preview` still on pricing. 2 3 4	Match ID to surface; GA vs preview is surface-specific.
Scene extension	10-second increments to 40 seconds cumulative; reads up to 10s of prior context (earlier models used the final second). Multi-turn via `previous_interaction_id`, or Files API upload. 1 2	Longer stories stay on one thread instead of stitching clips.
First / last frames	Two images plus a transition prompt generate continuous video between keyframes. 1 2	Camera moves become directed shots.
Resolutions	`360p`, `720p` (default), `1080p`, `4k` (1080p/4K upscaled). Google: 360p up to 60% faster and ~1/3 the 720p cost. Aspects: `16:9`, `9:16`. 1 2	Draft cheap at 360p; finish only the keeper.
Video references	Up to three clips, ≤3s each; reference audio ignored. 2	Short likeness or motion cues with text/image.
Access	AI Studio, Gemini API, Agent Platform; Flow and Gemini-app scene extension for AI Plus/Pro/Ultra. Cited integrations: Adobe Firefly, Figma Weave, Runway. Hosted only. 1	Same-day consumer and developer paths.
Price	Paid tier only. Tokens: input $1.50/1M; text out $9; video out $17.50. Blog $/s ladder: 360p $0.03, 720p $0.10, 1080p $0.15, 4K $0.30. Docs: 5,792 tokens/s at 720p ≈ $0.10/s. 1 4	Finish resolution drives most of the bill.
Hard limits	Upload edit/extend ≤10s input; append-only extension. Upload edit/extend blocked in EEA, Switzerland, UK (model-generated multi-turn still allowed). No free tier; no system instructions, temperature, top_p, or negative-prompt fields; no voice editing or audio reference uploads; SynthID on outputs. Model card flags consistency under edits, complex motion, and perfect on-screen text. 2 5	Region routing and clip length before productizing.

Explicit video_config.task values include text_to_video, image_to_video, reference_to_video, edit, and extend. Multi-turn edits use the Interactions API so you do not re-upload the prior clip. 2

Pricing ladder

Per-second USD pricing for Gemini Omni 1.1 Flash, Gemini Omni Flash, Veo 3.1 Lite, and Veo 3.1 Fast across 360p to 4K — Google's launch table prices Omni 1.1 Flash at $0.03/s (360p), $0.10/s (720p), $0.15/s (1080p), and $0.30/s (4K). Prior Omni Flash lists 720p only at $0.10/s. Veo 3.1 Lite is $0.05/$0.08 (720p/1080p); Veo 3.1 Fast is $0.10/$0.12/$0.30 (720p/1080p/4K). 1

Why this matters

Omni now supports a draft-extend-finish loop in one API: cheap 360p exploration, keyframe-directed shots, up to 40 seconds of extended narrative, then 1080p or 4K on the keeper. For a creative tool or media editor, run a 360p matrix on gemini-omni-1.1-flash — text-to-video, first/last-frame transitions, one multi-turn extend — then price the same prompts at 720p and 4K against your target length. Plan around the 10-second upload cap, end-of-clip-only extension, EEA/UK/CH upload-edit rules, and paid-tier-only access before promising in-product editing on customer footage. 1 2 4