Key takeaways
- Google announced Gemini Omni 1.2 Flash on November 5, 2026 and made gemini-omni-1.2-flash generally available on the paid Gemini API tier and across Gemini for business.
- The model generates synchronized video and native audio for any whole-second duration from 3 to 15 seconds. It supports stateful conversational edits of clips up to 15 seconds and extensions of 3–10 seconds per turn, up to 30 seconds total.
- Google prices standard output at $12 per million tokens: approximately $0.023 per second at 360p, $0.07 at 720p, $0.104 at 1080p with upsampling and $0.209 at 4K with upsampling. A 15-second 720p clip is listed at about $1.04 before retries.
- gemini-omni-1.1-flash remains available for a four-week transition. On December 3, 2026, Google says existing 1.1 API requests will automatically route to 1.2 with no code change and use the lower 1.2 rates.
- Treat the reroute as a production migration, not a billing-only change. Freeze representative prompts and references now, compare outputs and accepted-clip cost, then set quality and rollback gates before December 3.
Gemini Omni 1.2 Flash became generally available on November 5
Google announced Gemini Omni 1.2 Flash on November 5, 2026 and made it generally available across the Gemini API and Gemini for business. Developers call the paid API model as gemini-omni-1.2-flash. Google also says the release is rolling out through Flow, Google Vids, the Gemini app for Google AI Ultra subscribers and Google Ads, but those product routes have their own plan and rollout conditions.
The central change is a larger editable unit: Omni 1.2 generates synchronized video and native audio for 3 to 15 seconds, up from the ten-second native ceiling Google describes for Omni 1.1. The API accepts text, image and video inputs. Google positions the update around faster generation, stronger reference adherence, more consistent characters and objects, more natural dialogue and steadier camera motion; AccessAllGPT did not reproduce those quality claims.
Fifteen seconds comes with mode-specific limits
The response format accepts any whole-second duration from 3 through 15 seconds and defaults to ten seconds. Single-image generation can use the full range, but first-to-last-frame interpolation is limited to 3–10 seconds. Clips lasting 11–15 seconds accept one reference video of up to three seconds; 3–10 second clips accept up to three reference videos. All durations can use as many as six reference images.
That distinction matters when translating a storyboard into an API request. A 15-second output is not automatically compatible with every control available to a shorter clip. Teams should encode duration, reference count and generation mode as validation rules rather than rely on prompt text: the API returns an invalid-argument error when two-frame interpolation requests more than ten seconds.
Conversational edits preserve context, but every turn creates a new video
Through the Interactions API, a follow-up request can reference the previous interaction and ask for a focused change such as different lighting, weather, background or aspect ratio. Google says the model remembers prior video context and preserves elements the prompt does not mention. Generated and uploaded videos up to 15 seconds can be edited, subject to regional and content restrictions. Each turn produces a new video, so editing still consumes generation time and output tokens.
Extension appends 3–10 seconds at the end of an uploaded or model-generated clip and can continue through multiple turns to 30 seconds total. The model uses the final ten seconds as context. It cannot prepend or insert footage in the middle, and an uploaded clip with someone already talking cannot be extended with additional dialogue. Multi-turn extension of model-generated speech is supported.
The 720p rate falls to about seven cents per generated second
Google lists Gemini Omni 1.2 Flash only on the paid Gemini API tier. Standard output costs $12 per million tokens. The pricing page maps resolution to token use and effective rates: 1,931 tokens per second and about $0.023 at 360p; 5,792 tokens and about $0.07 at 720p; 8,688 tokens and about $0.104 at 1080p with upsampling; and 17,376 tokens and about $0.209 at 4K with upsampling.
At the listed rates, Google estimates a ten-second 720p output at about $0.70 and a 15-second output at about $1.04. Those are generated-clip prices, not accepted-deliverable prices. A three-take workflow plus two conversational edits can multiply output cost even when only one final clip ships. Measure rejected outputs, edit turns and upscaling alongside the nominal per-second rate.
Existing Omni 1.1 traffic will move automatically on December 3
Google’s developer guide gives gemini-omni-1.1-flash a four-week transition window. Effective December 3, 2026, all API requests sent to that identifier will automatically route to gemini-omni-1.2-flash with no code change and will be billed at the lower 1.2 rates. That avoids an immediate identifier migration, but it also means unchanged applications can receive a different model.
Before the cutoff, save a frozen set of production prompts, source media and accepted 1.1 outputs. Run the same cases through 1.2 and compare identity, product detail, text, speech, timing, camera motion, safety outcomes, latency and total cost per accepted clip. If the application must preserve 1.1 behavior beyond December 3, Google’s current documentation does not promise that option; ask Google for an explicit contract rather than assume the old alias will remain pinned.
Regional, media and provenance constraints still shape deployment
Editing or extending uploaded video is unavailable in the European Economic Area, Switzerland and the United Kingdom, although Google says editing and extension of model-generated video remain supported in available regions. Uploaded audio references are unsupported, audio in reference videos is ignored, voice editing is unsupported and reasoning across multiple videos is not supported. Large outputs should use URI delivery rather than inline payloads.
Google says all generated videos contain an imperceptible, programmatically detectable SynthID watermark and that safety filters apply to prompts and outputs. AccessAllGPT did not test detection or watermark durability. Google also names fast sports action, rapid script changes and complex physics as areas still being improved. These caveats should stay in acceptance tests instead of being obscured by the launch demos.
What builders and buyers should do next
New projects should prototype with the exact gemini-omni-1.2-flash identifier and a representative set of 3-, 10- and 15-second briefs. Separate tests by generation mode, reference count, resolution and edit path. Record every output and edit turn, total generated seconds, latency, errors, human acceptance and final cost. Keep rights and consent records for every source image or video and require a person to approve public-facing output.
Existing Omni 1.1 users should treat December 3 as a dated migration gate. Compare the two routes before then, update cost forecasts for the lower rates, validate regional behavior and set a rollback or non-generative fallback for failed cases. Wait when a workflow needs unsupported voice editing, multi-video reasoning, mid-clip insertion or a verified quality guarantee that Google’s launch material does not provide.
Copy-ready Gemini Omni 1.2 migration record
Complete one record per production video workflow before starting new traffic or allowing the December 3 automatic reroute.
Entries stay in this browser tab and are not submitted to AccessAllGPT. Blank responses are copied as [Unresolved].
Use case, audience, owner, new integration or 1.1 migration; trial, constrain, approve, wait or reject; review and expiry dates.
gemini-omni-1.2-flash or current 1.1 alias, API project, paid tier, region, SDK version, quota and December 3 transition exposure.
Prompts, people, voices, products, brands, reference images and videos; ownership, consent, license, retention and prohibited transformations.
Mode, duration, aspect ratio, resolution, native audio, references, interpolation, stateful edit or extension, delivery method and unsupported controls.
Representative briefs and edge cases covering identity, product fidelity, text, speech, timing, camera movement, physics, rapid action and safety blocks.
Accepted outputs, rejection reasons, edit turns, generated versus accepted seconds, latency, failures, regional behavior and reviewer agreement.
Output token rate, resolution, clip duration, retries, edits, upsampling, storage, review labor and cost per accepted deliverable.
Human approval, SynthID check, disclosure, content policy, timeout, traffic ceiling, rollback or non-generative fallback and incident owner.
Primary sources
Browse the publication-wide evidence index →
- Gemini Omni 1.2 Flash: Fast and cost-effective with longer clipsGoogle / Google DeepMind · Reviewed: November 5 announcement; general availability; 15-second generation; conversational editing; reference media; speech and camera-control claims; product access; SynthID; known weak areas; migration and pricing links · Retrieved · Supports: Google announced Gemini Omni 1.2 Flash on November 5, 2026, made it generally available across the Gemini API and Gemini for business, and described 15-second video, multi-turn editing and a 30% lower 720p output price. Quality statements and examples are vendor claims.
- Generate and edit videos with Gemini Omni FlashGoogle AI Studio documentation · Reviewed: Model ID and transition notice; inputs and native audio; duration and resolution controls; reference limits; stateful editing; extension; URI delivery; regional and media limitations; SynthID and safety filters · Retrieved · Supports: The live documentation identifies gemini-omni-1.2-flash and documents 3–15 second generation, up to six reference images, stateful editing, 3–10 second extensions up to 30 seconds total, and material input, region and editing constraints.
- Gemini Developer API pricing — Gemini Omni 1.2 FlashGoogle AI Studio documentation · Reviewed: Paid-tier availability; standard output-token rate; resolution-specific token rates and effective per-second examples; Omni 1.1 transition pricing · Retrieved · Supports: Google lists no free API tier for Gemini Omni 1.2 Flash and prices output at $12 per million tokens. It maps that to about $0.023 per second at 360p, $0.07 at 720p, $0.104 at 1080p with upsampling and $0.209 at 4K with upsampling.
Limitations
This is a first-party launch and documentation review, not hands-on video-model testing or an independent benchmark. AccessAllGPT did not authenticate to Google AI Studio, Gemini, Flow, Vids, Ads, Gemini for business or Google Cloud; call gemini-omni-1.2-flash; compare it with Omni 1.1 or another model; generate, edit, interpolate, extend, upscale or download a video; test reference limits or regional restrictions; measure latency, consistency, speech, prompt following, text rendering, camera motion, physics, safety filters or accepted-output cost; inspect token accounting; verify Provisioned Throughput; or detect SynthID. Google controls all cited sources. Its quality, speed and consistency descriptions, demos and customer quotations are vendor evidence, not independent results. Product availability, rates, quotas and automatic-routing behavior can change.
Disclosures
AccessAllGPT received no Google access, API credits, subscription, private demo, test output, briefing, benchmark data, review, payment or compensation for this article. Google and Google DeepMind did not sponsor, review or endorse it. AccessAllGPT did not test, score, rank or certify Gemini Omni 1.2 Flash. AccessAllGPT Research is operated by NeuralArc, is independent, and is not affiliated with Google or any company quoted in the launch post. Publication-wide relationships are listed on the disclosures page.
