Overview
Gemini Omni 1.2 Flash is the November 5, 2026 update to Google's video generation and conversational editing model. For creators making product clips, try-ons, and social ads, the clearest change is a longer single generation: up to 15 seconds instead of 10. Google also lowers 720p API video output pricing to about $0.07 per second and reports stronger subject consistency.
For existing API users, the upgrade has a deadline: requests to the 1.1 model ID will automatically use 1.2 from December 3, 2026. Longer single clips do not raise every workflow limit; interpolation remains capped at 10 seconds, and the current API extension ceiling is 30 seconds total. This guide is based on official documentation and selected Google demos, without ToolWorthy hands-on testing.
What's New
Generate and edit up to 15 seconds
Version 1.2 increases the native generation and editing limit from 10 to 15 seconds. Duration can be set in whole seconds from 3 to 15, with 10 seconds as the API default. This can accommodate a short product reveal or one ad scene without joining separate generations.
That improvement has boundaries: first-and-last-frame interpolation supports only 3–10 seconds. A single starting image can be used for generation up to 15 seconds. For extended sequences, the Gemini API currently allows 3–10 seconds per continuation and 30 seconds total.
Improve subject consistency with up to six reference images
Google reports tighter adherence to products, clothing, characters, and scene details. The API supports up to six reference images, with up to three images per character for up to two characters. These controls are relevant when a clip must preserve a product's identity across motion.
A longer clip also changes video-reference capacity: 11–15 second generation supports one reference clip of up to three seconds; 3–10 second generation supports up to three. Reference videos guide new generation, not editing or extension, and their audio is ignored. Google's selected examples do not establish a measured success rate for your assets.
Reduce video output costs
The Gemini API video-output rate falls from $17.50 to $12 per million tokens. Google's rounded 720p figure is about $0.07 per second, advertised as 30% less than 1.1. Input and text-output token prices stay at $1.50 and $9 per million respectively.
Conversational editing already existed in 1.1. The 1.2 improvement is its application across longer clips and Google's reported gains in consistency, speech, and camera control; this is not the debut of video editing by conversation.
Compared With Previous Version
| Decision point | Omni 1.1 Flash | Omni 1.2 Flash |
|---|---|---|
| Single generated clip | 3–10 seconds | 3–15 seconds; default 10 |
| Approximate 720p API video output | $0.10/second | $0.07/second |
| API video-output token price | $17.50 per million | $12 per million |
| Reference images | Reference-guided generation and editing | Up to six, subject to per-character limits |
| First-and-last-frame interpolation | Up to 10 seconds | Still up to 10 seconds |
| Extended sequence | Earlier 1.1 documentation described up to 40 seconds | Current 1.2 guide caps extension at 30 seconds total |
| API lifecycle | Separate model during the transition window | Receives 1.1 traffic automatically from December 3 |
The longer native clip is useful when the scene fits in 15 seconds. A workflow that relied on the older 40-second extension claim needs a regression check rather than assuming the new model preserves that limit. The Gemini Omni tool guide covers the overall product; the 1.1 release page preserves its historical launch context.
Pricing Changes
The following are Gemini API paid-tier rates, not a Flow credit schedule or a Gemini subscription price. There is no free API tier for this model.
| API component | Omni 1.2 Flash rate |
|---|---|
| Text, image, and video input | $1.50 per million tokens |
| Text output | $9 per million tokens |
| Video output | $12 per million tokens |
| 720p video output | Approximately $0.07/second |
| 15-second 720p clip, video output only | Approximately $1.04 |
Google calculates 720p output at 5,792 tokens per second: 15 × 5,792 × $12 / 1,000,000 = $1.04256. Inputs, text output, additional edits, and retries can increase the total. The pricing page also lists about $0.023/second for 360p, $0.104/second for upscaled 1080p, and $0.209/second for upscaled 4K. Choosing a draft at 360p before final delivery can reduce the cost of discarded concepts.
Migration and Access
Developers can explicitly select gemini-omni-1.2-flash in the Gemini API and prototype in Google AI Studio. Google says that from December 3, 2026, requests using gemini-omni-1.1-flash automatically route to 1.2 and use its lower rates without requiring a model-ID code change. This is a scheduled behavior change, not a promise of identical outputs.
Before that date, run representative prompts through 1.2, including interpolation, longer clips, reference-video counts, and any workflow needing more than 30 seconds. Check the generated product details, motion, dialogue, and cost before switching a production workload. Google has not provided a version-isolated quality guarantee for those prompts.
Creator access varies by surface. The launch announcement names Flow, Google Vids, and a Gemini app rollout to Google AI Ultra subscribers. It describes 15-second generation in Google Vids at no additional cost with an eligible Google account or Workspace subscription; that does not make API use free. Google Cloud's launch page still describes Gemini for business as coming soon in its access section, so verify availability in the intended product rather than assuming the rollout has reached every account.
Who Should Upgrade / Who Should Wait
Evaluate 1.2 now if:
- Your ad, product reveal, or try-on needs 11–15 seconds in one generation.
- Your workflow benefits from up to six image references and lower per-second video-output costs.
- You maintain a 1.1 API integration and need to prepare for the December 3 routing change.
Keep an evaluation stage if:
- You depend on extensions longer than the current 30-second ceiling or two-frame interpolation longer than 10 seconds.
- Your product needs exact text, garment details, or perfect identity consistency; vendor demos are not a pass rate for your inputs.
- You need uploaded audio references or voice replacement in the Gemini API, where these remain unsupported.
Known Limitations
The Gemini API does not support uploaded audio inputs or audio references, voice editing, or audio from reference videos. Google Cloud separately says audio references require allowlisting; those Cloud capabilities must not be treated as standard Gemini API features.
Editing or extending uploaded video is unavailable in the EEA, Switzerland, and the UK, while editing or extending model-generated videos is supported. Outputs at 1080p and 4K use upscaling. SynthID watermarking remains mandatory. Google identifies fast sports action, rapid script changes, and complex physics as areas still being improved.
