On November 5, 2026, Google announced the general availability of Gemini Omni 1.2 Flash, its AI video generation and editing model. The maximum length for standard video generation and editing has been extended from 10 seconds to 15 seconds, and the output price for 720p video has been lowered from about $0.10 per second to about $0.07 per second. Google describes this as a price cut of roughly 30%. In the Gemini API, a 15-second 720p video works out to an output cost of about $1.04. However, inputs and other items are billed separately, and the available length and conditions differ for generating, editing, and extending videos. To assess real production costs, you need to consider not only the output price of a single run but also how many revisions are needed to reach a finished result.
Support for videos up to 15 seconds, with improved visual consistency
With Gemini Omni 1.2 Flash, you can specify a length of 3 to 15 seconds in one-second increments for standard video generation and editing. Version 1.1 was capped at 10 seconds, so a single run can now handle videos that are 5 seconds longer. You can also instruct the model to change the background or lighting of an existing video and edit the entire clip, up to 15 seconds.
Google says this update also improves the model's ability to preserve the appearance of referenced people and products. By using not only images but also short motion clips of up to 3 seconds and audio cues, users can get closer to the footage they intend, according to the company. Google also says it has improved audio and lip synchronization in scenes with multiple speakers, as well as the rendering of continuous camera movement. However, the published samples were selected by Google, and they are not the result of independent verification of, for example, how often the model produces footage that matches the instructions.
That said, the ability to edit videos through iterative dialogue is not new in itself. Earlier versions also offered video extension and video generation that connects two images, and they supported a workflow of creating drafts at 360p or 720p and then upscaling to 1080p or 4K. In the Gemini API, version 1.1 became generally available on August 27, 2026. Version 1.2 can be seen as an update that keeps the existing production workflow while extending the length of video the model can handle, improving the fidelity of reference material, and lowering output prices.
Up to 15 seconds for generation and editing, with separate limits for extension
With Gemini Omni 1.2 Flash, the length of video the model can handle varies by type of operation. Standard generation and editing support up to 15 seconds, whereas interpolation, which generates the motion between a specified first and last image, is limited to 10 seconds. For extension, which adds a continuation to an existing video, there are separate limits on the length that can be added in a single run and on the total length of the video after extension.
The limits for each operation, based on the Gemini API model specifications, are as follows.
| Operation | Description | Length in the Gemini API |
|---|---|---|
| Standard generation | Generates video from text or reference material | 3–15 seconds of output |
| Editing | Modifies the content of an existing video | Up to 15 seconds of input |
| Interpolation between two images | Generates intermediate footage from a specified first and last image | 3–10 seconds of output |
| Extension | Generates a continuation of an existing video | 3–10 seconds per run, up to 30 seconds in total |
For extension, the model references the final 10 seconds of the existing video to generate the continuation. However, Google Cloud has different conditions for linking clips during extension, so the Gemini API specifications cannot be applied there as is.
For audio as well, it is necessary to distinguish between generating audio together with video and using input audio as reference material. Google Cloud's announcement states that the audio reference feature requires prior approval. At the same time, Cloud's specification table also lists audio input as unsupported, so not every user can freely supply audio material. Google further says it is still working to improve the rendering of fast sports movements, abrupt scene changes, and complex physical phenomena.
About $1.04 for 15 seconds at 720p: how pricing compares with rival models
With the general availability of Gemini Omni 1.2 Flash, the video output price in Gemini API Standard has been lowered from $17.50 to $12 per million tokens. Because 720p video consumes 5,792 tokens per second, the output cost for 15 seconds can be calculated as follows.
[ 5,792 \times 15 \times \frac{12}{1,000,000} = 1.04256\ \text{USD} ]
In other words, the output cost of a 15-second 720p video is about $1.04. This differs slightly from $1.05, which is the published approximate per-second price of $0.07 multiplied by 15, because of rounding. In addition, the input price is $1.50 per million tokens and text output is $9 per million tokens, and these are calculated separately from the video output price.
How does Omni 1.2 Flash's pricing compare with other video generation models? Based on Google's official pricing page and Runway Dev's pricing page, we converted the standard listed prices as of November 6, 2026 into US dollars per second of output.
| Model | Service | Audio conditions / basis of comparison | Output price (USD/sec) |
|---|---|---|---|
| Gemini Omni 1.1 Flash | Gemini API | 720p video output | about 0.10 |
| Gemini Omni 1.2 Flash | Gemini API | 720p video output | about 0.07 |
| Gen-4.5 | Runway Dev | Not compared as audio-generating | 0.12 |
| Gen-4 Turbo | Runway Dev | Not compared as audio-generating | 0.05 |
| Kling 3.0 Standard | Runway Dev | With / without audio | 0.13 / 0.09 |
| Veo 3.1 Fast | Runway Dev | With audio | 0.15 |
| Veo 3.1 | Runway Dev | With audio | 0.40 |
Runway Dev prices are converted at 1 credit = $0.01; for example, Gen-4.5's 12 credits per second equals $0.12 per second. These are a separate pricing system from the credits granted through Runway app subscriptions, and taxes may be added.
The table shows that Omni 1.2 Flash's roughly $0.07 per second is cheaper than Gen-4.5 and Veo 3.1, among others, but not lower than Gen-4 Turbo's $0.05 per second. However, the services, resolutions, and audio generation conditions differ, so this is not a price comparison under identical conditions. In addition, generation quality, processing speed, and supported video lengths have not been evaluated on a common basis, so this table alone cannot determine which model offers the best cost performance.
Omni 1.2 Flash's pricing also varies by resolution. The video output charge is about $0.023 per second at 360p and about $0.07 per second at 720p, and upscaling to 1080p costs about $0.104 per second, while upscaling to 4K costs about $0.209 per second. If you iterate on prototypes at low resolution and finish at high resolution, you need to account for the cost of each step.
Furthermore, what drives actual video production costs is not the per-second output price alone. For example, when changing only the background while keeping a product's appearance intact, improved consistency in the generated results could reduce how often the entire video has to be remade. If revisions decrease, production costs could fall by more than the output price cut alone. However, Google has not quantitatively shown how much revision counts, inspection time, or final production costs will be reduced.
Old API to migrate automatically on December 3, with usage conditions varying by app
Starting December 3, 2026, Google plans to automatically redirect requests to the legacy Gemini API model ID gemini-omni-1.1-flash to version 1.2. After the migration, 1.2 pricing will apply, so developers still using the old model ID will be affected. However, this automatic migration applies to the Gemini API, and the same conditions do not necessarily apply to environments such as Google Cloud or Runway.
Omni 1.2 Flash is also rolling out across Google's apps. Google Vids will offer 15-second video clips at no additional charge to users with Google accounts and Workspace. In the Gemini app, the rollout begins with Google AI Ultra users, and the model can be used in Flow for directing and editing video and in Google Ads for producing ad creative. Meanwhile, the Gemini API has no free tier, and availability for "Gemini for business," announced by Google Cloud, is planned for the future.
However, the fact that some apps can be used for free does not mean every Omni 1.2 Flash feature is free. Android Authority points out that Google has not made clear whether all of 1.2's features are available in the free version of Flow. In Google Vids as well, it is unclear how far improvements beyond 15-second video generation are reflected.
Developers currently using Gemini Omni 1.1 Flash should check whether their current prompts and production workflows still work as intended on 1.2 ahead of the December 3 automatic migration. In particular, the video length required, the editing operations used, and the fidelity of reference material matter, as does the number of attempts needed to reach acceptable quality. The extension to 15 seconds and the lower output price are clear improvements, but whether they translate into real production cost savings will depend on the stability of the generated results and editing efficiency.