Gemini Omni Flash price: the model replacing the Veo 3.1 previews costs about $1 for 10 seconds
On Google’s Gemini API price list, Gemini Omni Flash charges $17.50 per million video output tokens. A second of 720p counts as 5,792 tokens, so it works out to about $0.10 a second, and there is no free tier. It makes 3 to 10 seconds per run in 9:16 or 16:9. Google lists Omni as the replacement for the Veo 3.1 previews that shut down on October 22.
Gemini Omni Flash (gemini-omni-1.1-flash) is the video generation and editing model Google released as a stable version on August 27, 2026. It makes videos from text and images, edits them through conversation and extends them at the end. Google lists Omni as the replacement for the three Veo 3.1 previews that shut down on October 22.
Prices and specs below come from the Gemini API price list, the model page, the Omni docs and the deprecations page. We added what Sera Studio measured with free model lookups. The Kling 4.0, Seedance 2.5 and Veo prices we compare against come from our earlier price guides.
Gemini Omni Flash prices on one sheet
Gemini Omni Flash
- 720p per second · USD
- ~$0.10Google price list, token estimate
- 1080p per second
- Not listed
- Length per run
- 3–10 s
- Sound
- On by default
Kling 4.0
- 720p per second · USD
- $0.14Kling API, preview price
- 1080p per second
- $0.21
- Length per run
- 3–30 s
- Sound
- Always, stereo
Seedance 2.5
- 720p per second · USD
- $0.231BytePlus, checked Oct 23
- 1080p per second
- $0.569
- Length per run
- 4–30 s
- Sound
- Supported
A second of 720p costs about $0.10, and there is no free tier
The Gemini API price list bills Omni’s video output in tokens (Gemini Developer API pricing).
- Video output costs $17.50 per million tokens, and a second of 720p counts as 5,792 tokens. Google puts that at about $0.10 a second; the exact figure is $0.1014.
- Input costs $1.50 per million tokens for text, images, video and audio alike. Text output costs $9.00.
- The free tier column reads “Not available”. The list says paid-tier data isn’t used to improve Google’s products.
- The list gives no tokens-per-second figure for 360p, 1080p or 4K.
| 720p length | Output tokens | Price |
|---|---|---|
| 3 s | 17,376 | $0.30 |
| 5 s | 28,960 | $0.51 |
| 8 s | 46,336 | $0.81 |
| 10 s | 57,920 | $1.01 |
A 10-second 720p clip uses 57,920 tokens and costs about $1.01.
3 to 10 seconds per run, in 9:16 or 16:9
These specs come from the model page and the Omni docs (Gemini Omni Flash, Generate and edit videos with Gemini Omni Flash).
- Output runs 3 to 10 seconds at 24 fps. Resolutions are 360p, 720p (the default), 1080p and 4K, and the last two are upscaled.
- Aspect ratio is either portrait 9:16 or landscape 16:9.
- One model handles text-to-video, image-to-video, interpolation between a first and last frame, subject reference images, conversational edits, edits to your own uploads and extension.
- It generates sound by default.
- Every video it makes carries a SynthID watermark.
Extension only adds 3 to 10 seconds at the end
- Extension appends a 3 to 10 second continuation to the end of a video. It can’t add to the start or the middle.
- Uploaded videos must be 10 seconds or shorter to edit or extend. Videos Omni made itself are extended by continuing the conversation.
- You can’t add new dialogue when extending an uploaded video of someone talking.
Only English is fully supported, and some features are off limits
- The Omni docs say English is fully supported and other languages haven’t been evaluated, so results may vary.
- Voice editing and audio references aren’t supported. Video references are limited to three clips of up to 3 seconds each.
- You can’t upload or edit images of certain recognizable people, and YouTube videos can’t be used as a source.
- Editing or extending uploaded videos isn’t available in the European Economic Area, Switzerland or the UK.
The Veo 3.1 previews ended on October 22
- The deprecations page lists October 22 as the shutdown date for the Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite previews, with gemini-omni-1.1-flash as the replacement for all three (Gemini deprecations).
- Omni’s own preview, gemini-omni-flash-preview, ended the same day. It had come out on June 30, 2026.
- The stable Veo 3.0 and 2.0 models shut down on June 30, 2026. The page now lists their replacements as Omni or the stable Veo models on Google Cloud’s Gemini Enterprise Agent Platform.
- On October 25 the Gemini API price list had no Veo section.
- On the price list we checked on October 22, before the shutdown, a second of 720p cost $0.10 on Fast and $0.05 on Lite (see our Veo 3.1 Fast vs Lite guide). Omni costs about the same as Fast and twice as much as Lite.
What we measured
On October 25 we looked up Omni and the retired previews with the Gemini API’s free model lookup. We didn’t generate any video.
- gemini-omni-1.1-flash returned an input limit of 131,072 tokens and an output limit of 65,536.
- An output limit of 65,536 tokens covers about 11.3 seconds of 720p, so the 10-second cap (57,920 tokens) fits inside it.
- gemini-omni-flash-preview and the three Veo 3.1 previews, all past their shutdown date, still answered the lookup. We didn’t test whether they still take generation requests.
| Model id | Input limit | Output limit | Shutdown |
|---|---|---|---|
| gemini-omni-1.1-flash | 131,072 | 65,536 | None |
| gemini-omni-flash-preview | 131,072 | 65,536 | Oct 22 |
| veo-3.1-fast-generate-preview | 480 | 8,192 | Oct 22 |
| veo-3.1-lite-generate-preview | 480 | 8,192 | Oct 22 |
Which one to pick
The rest of this section is our analysis, not Google’s documentation.
A 10-second 720p clip costs about $1.01 on Omni, $1.40 on Kling 4.0 and $2.31 on Seedance 2.5 (Kling checked Oct 25, Seedance Oct 23; see our Kling 4.0 price guide). For short vertical clips you refine over several tries, Omni is the cheapest of the three. For a single clip longer than 10 seconds, Kling 4.0 or Seedance 2.5, which both go up to 30 seconds, fits better. If your video has dialogue in a language other than English, run a short test first. If you need to know the 1080p price today, Kling 4.0 and Seedance 2.5 publish one.
What we couldn’t confirm
- The price list gives no price for 360p, 1080p or 4K, and no token count for input images or video.
- The model page lists a 1,048,576-token context window, but the lookup returned an input limit of 131,072.
- We didn’t check whether the retired previews still accept generation requests.
- We haven’t made any videos with Omni, so we didn’t measure quality or speed.
- We didn’t cover the plans or usage limits in the Gemini app.
- This article was written by Sera Studio. Sera Studio’s Motion Studio offered Veo 3.1 Fast and Lite through the Gemini API until the previews shut down, and doesn’t offer Omni yet.


