Gemini Omni 1.2 Flash pricing: $0.07 a second, so here is what a finished 15-second clip really costs

A finished 15-second Gemini Omni 1.2 Flash clip costs about $1.04 at 720p, $1.56 at 1080p and $3.13 at 4K on the Gemini API, based on Google's published per-second rates as of November 2026. Drafts are where you save: a 15-second 360p draft costs about $0.35, a third of the 720p price.

Below: what each resolution costs per second and per clip, what a realistic draft-then-final workflow adds up to, and three billing questions Google's pricing page leaves open. Do this math before you make Omni your default model for short product or ad video.

What changed in Gemini Omni 1.2 Flash

Google made Gemini Omni 1.2 Flash generally available on November 5 with three changes that move the cost math:

• Lower price. Google says the price is down 30% from Omni 1.1, to $0.07 per second of 720p video.
• Longer clips. One generation now runs up to 15 seconds, up from 10 seconds with Omni 1.1, and you set the length in one-second steps from 3 to 15 seconds.
• Draft, then upscale. You can create drafts at 360p or 720p, then upscale the best ones to 1080p or 4K.

The 30% is real but rounded. On the Gemini API pricing page, video output for Omni 1.1 Flash was $17.50 per million tokens and Omni 1.2 Flash is $12.00, a cut of about 31%.

Gemini Omni 1.2 Flash pricing is per token, not per second

Google does not bill Omni by the second. The pricing page bills output tokens at $12.00 per million for video, and converts that into seconds with a fixed token rate per resolution: 5,792 tokens per second of 720p video. The $0.07 headline is that conversion: 5,792 × $12.00 ÷ 1,000,000 = $0.0695 per second.

The same footnote lists the other resolutions. Here is every rate, multiplied out to the clip lengths people actually use:

The token rates follow a clean pattern. 360p costs a third of 720p, 1080p costs 1.5 times 720p, and 4K costs exactly three times 720p. So a 15-second 4K clip (17,376 × 15 = 260,640 tokens, or $3.13) costs the same as three 15-second 720p clips.

The pricing page lists no separate charge for audio, which the Omni docs say the model generates by default, synchronized with the video. Input is billed too, at $1.50 per million tokens for text, images and video. A text prompt adds a fraction of a cent, but we did not find a published token count for reference images, so we have left input out of the totals below.

Gemini Omni 1.2 Flash vs 1.1: more footage for about the same money

The better comparison is per clip, not per second. Omni 1.1 Flash topped out at 10 seconds and cost about $0.10 per second at 720p, according to the pricing page, so its longest clip was roughly $1.01. Omni 1.2 Flash gives you a 15-second 720p clip for $1.04: half again as much footage for about 3 cents more.

If you are still calling Omni 1.1, you do not have to migrate by hand. The docs say that from December 3, 2026, requests to the gemini-omni-1.1-flash model route automatically to Omni 1.2 Flash and are billed at the lower 1.2 rates.

What a finished 15-second clip really costs

The realistic cost is the drafts you throw away plus the final render. Here is the arithmetic for four 15-second drafts and one final, using only the published rates:

The resolution you draft at matters more than the resolution you finish at. Moving drafts from 720p to 360p saves $2.78 per finished clip, while dropping the final from 4K to 1080p saves $1.56. If 360p is enough to judge framing, motion and the product shot, draft there.

Skip 4K if the clip is headed for a social feed or an ad slot that plays it smaller. At three times the 720p price, 4K is worth it only when the delivery format actually shows the difference.

Three billing questions the pricing page does not answer

We read the pricing page and the Omni docs line by line. Three things that affect your budget are not stated, as of November 6, 2026:

1. How upscaling a draft is billed. In the API docs, resolution is a parameter you set on the generation request, and 1080p and 4K are marked as upscaled outputs. Neither page describes a separate upscale call on an existing draft or its price, so the tables above budget the final as a full render at the listed rate.
2. Whether conversational edits cost a full clip. The docs let you edit a generated clip of up to 15 seconds in follow-up turns. The pricing page bills output tokens and says nothing about edits, so the safe assumption is that each edited clip is billed like a new one.
3. How extensions are billed. Extension adds 3 to 10 seconds per turn, up to 30 seconds in total. The pricing page does not say whether you pay for the new seconds or the whole returned video.

Until Google says otherwise, put a cap on edit and extension turns in any script that calls the API in a loop.

Where you can use Gemini Omni 1.2 Flash, and what you pay there

Google's announcement lists these surfaces, and the price depends on which one you use:

• Gemini API and Google AI Studio. Pay per token as above. The pricing page shows no free tier for Omni 1.2 Flash in the API.
• Google Cloud. The Cloud GA post says the model is available through the Cloud console and an API, with Provisioned Throughput, which reserves capacity for production workloads. The Gemini API docs say Provisioned Throughput is not supported there, so that is a reason to pick Cloud. The AI Studio pricing page warns that prices may differ on Google Cloud's platform, so check its own rate card.
• Flow. Available for directing scenes and multi-turn edits. Our guide to editing video with AI in Google Flow covers how that app works.
• Google Vids. Google says Vids makes 15-second clips at no additional cost with a Google account or Workspace subscription.
• Gemini app. 15-second videos are rolling out to Google AI Ultra subscribers.

For a handful of clips a month, Vids is the cheapest route, though Google's post gives no usage limits, so check what your plan allows. The API makes sense once you generate at volume or build video into your own product.

What Google does not claim yet

Google says Omni 1.2 Flash keeps characters, objects and scenes more consistent from first frame to last, with clearer lip-sync and fewer unwanted jump cuts. It publishes no benchmark scores against 1.1, so quality is something to test on your own prompts. Google also says it is still working on fast sports action, rapid script changes and complex physics.

The docs add limits: editing or extending uploaded videos is not available in the EEA, Switzerland or the UK, and audio inputs are not supported in the current API. Every output carries an invisible SynthID watermark.

The Cloud post names companies building with it, including Picsart, Photoroom, invideo and Runway. That shows production interest, not that Omni is right for every job.

Compare it before you commit a budget

Run your own cost check before you standardize: take one real brief, generate four 360p drafts and one final at the resolution you deliver, and compare the bill with the table above. Then price the same clip on the tools you already use.

Upfynd lists 574 AI image and video tools and 368 LLM and AI infrastructure products as of November 6, 2026, including API gateways that put many image and video models behind one key. If Gemini API costs are already on your radar, our breakdown of the Gemini API price changes on December 1 and January 1 is worth reading next. And if you build a video tool on top of Omni, you can list it on Upfynd for free.