Google's Omni 1.2 Flash Makes 15-Second Videos and Cuts the Price 30% to $0.07 a Second. Native Length Is Still a Clip, Not a Film.
Google DeepMind made Gemini Omni 1.2 Flash generally available on November 5 across the Gemini API and Gemini for business. Native clips move from 10 seconds on Omni 1.1 to 15, selectable from 3 to 15 seconds in one-second steps, with multi-turn conversational editing, stronger adherence to reference images and clips, and a 30% price cut to $0.07 per second at 720p, with upscaling to 1080p or 4K after drafting. Outputs carry a SynthID watermark. Google names its own weak spots: fast sports action, rapid script changes, and complex physics. All numbers are Google's. We read the DeepMind post, the Cloud post, the Vids post, and the model card.
By T.W. Ghost
Google DeepMind made Gemini Omni 1.2 Flash generally available today, November 5, 2026, on the Gemini API and Gemini for business, with the model id gemini-omni-1.2-flash. The changes from Omni 1.1: native clips up to 15 seconds instead of 10, with duration selectable from 3 to 15 seconds in one-second steps; multi-turn conversational editing across the full clip, so you can swap a product or a background, change the lighting, or change the aspect ratio without regenerating; stronger adherence to reference images, motion clips up to 3 seconds, and audio cues; and a 30% price cut to $0.07 per second at 720p.
The workflow Google describes is draft, edit, then upscale: generate at 360p or 720p, iterate in conversation, then upscale to 1080p or 4K. Every output carries an imperceptible SynthID watermark, and Provisioned Throughput is available. Google also lists where the model still falls down: fast sports action, rapid script changes, and complex physics. Every number here is Google's, and the comparison frames in the launch are Google's selection. We read the DeepMind announcement, the Google Cloud post, the Workspace post for Vids, and the model card. Here is what held up.
What Changed From Omni 1.1
| Omni 1.1 | Omni 1.2 Flash (GA November 5, 2026) | |
|---|---|---|
| Native clip length | 10 seconds | Up to 15 seconds; 3 to 15 in 1-second steps |
| Price, 720p | Higher | $0.07 per second; Google says 30% lower than 1.1 |
| Editing | Multi-turn conversational edits across the full clip | |
| References | Character, product, and style images; motion clips up to 3 seconds; audio cues (vendor) | |
| Dialogue | More natural multi-speaker dialogue and lip-sync (vendor) | |
| Pipeline | Draft at 360p or 720p, edit, upscale to 1080p or 4K | |
| Watermark | SynthID on all output | |
| Surfaces | AI Studio and Gemini API; Google Cloud console; Flow; Google Vids (15-second clips at no additional cost with a Google account or Workspace); Gemini app, rolling out to AI Ultra; Google Ads | |
| Known gaps (Google) | Fast sports action; rapid script changes; complex physics |
Fifteen seconds is the ceiling. This is a model for ads, product shots, and explainers, and Google's own surface list, Vids and Google Ads, says so. It is not long-form production, and the price is per second for a reason.
Per second still adds up. At $0.07 a second, a 15-second 720p clip lists at $1.05 before upscales, retries, and audio variants. Google's 30% is relative to Omni 1.1, not to the competition's promotions; fal's FLUX 3 Max launch rate, for instance, ends November 7.
Say which surface you mean. The API is metered. Vids includes 15-second clips at no extra cost. The Gemini app path is rolling out to AI Ultra. Three different prices for the same model, depending on the door.
The multi-turn edit is the cost lever. Paying to fix a lighting problem on an existing clip instead of paying to regenerate the whole thing is the change most likely to move a pipeline's bill, and it is the one to test first.
Verified vs Unconfirmed: The Scorecard
| Claim | Verdict |
|---|---|
| GA Nov 5 on Gemini API and Gemini for business | Vendor (Google blog; Cloud blog) |
| 15-second native clips; 3 to 15 in 1-second steps | Vendor |
| $0.07 per second at 720p; 30% below Omni 1.1 | Vendor |
| Draft, multi-turn edit, upscale to 1080p or 4K | Vendor |
| Reference adherence and dialogue improvements | Vendor, demo-selected |
| SynthID on all output; Provisioned Throughput | Vendor |
| Vids 15-second clips at no additional cost; Gemini app rolling to AI Ultra | Vendor (Workspace blog; Google blog) |
| Weak spots: sports, rapid script changes, physics | Vendor (Google's own list) |
| Independent quality comparison or win rate | None |
Who Should Use It
Re-price your pipeline if it assumed Omni 1.1's 10-second clips and old per-second rate, and move iterations onto the edit path instead of full regenerations.
Skip it for anything longer than a clip, and skip quoting consistency wins you have not measured yourself; the comparison frames are Google's picks.
The frontier is elsewhere: on the arenas, the hosted leader is still Seedance 2.5, the open-weight news is Pika H3 Flex, and DeepMind's own Genie 4 is a waitlist. Our model comparison has the current standings on the language models side by side.
Sources
- Google: introducing Gemini Omni 1.2 Flash
- Google Cloud: Gemini Omni 1.2 Flash generally available
- Google Workspace: Omni 1.2 Flash in Google Vids
- Google DeepMind: Gemini Omni 1.2 Flash model card
*Not sure whether a per-second video API or a hosted creative suite fits your output? Take the free 2-minute quiz and get matched. Then read about the video model that shipped weights instead of a meter: Pika H3 Flex.*