Omni Flash: $9 per 1M text tokens vs $17.50 per 1M video tokens

Google's pricing page lists Gemini Omni Flash at $9.00 per million text output tokens and $17.50 per million video output tokens. A 2,000-token text reply costs $0.018, which is about the price of 0.18 seconds of 720p video, so text is a rounding error next to the video itself.

The video figure is where the money is: 5,792 tokens a second at $17.50 per million is about $0.10 a second.

The two meters

Video: 5,792 tokens per second of 720p video, times $17.50 per million, is $0.1014 a second. A 10-second clip is 57,920 tokens, or about $1.01.

Text: $9.00 per million tokens is $0.009 per thousand. If a request returns a 500-token description, that is $0.0045. Even 20,000 text tokens would be $0.18.

Why it matters

If you budget Omni calls by the per-second rate alone, you will be within a few percent of the Google invoice for ordinary prompts. The text meter only matters if your workflow asks for long written replies alongside video, as in a multi-turn chat that returns commentary.

We did not verify from the page whether a plain video request returns any billed text at all. The sum above is an upper bound if it does.

A planning shortcut

Use the video figure only. Take the seconds you will generate, multiply by $0.1014 for Google-direct 720p, and add a few percent if you also ask for long text. That will land within a few percent of the invoice for most plain video requests.

If you ever see a bill well above that, look first at retries and failed jobs, not at text tokens.

On Sume

Sume does not bill tokens for Omni. Its docs say it bills the provider list times 1.25 for each output second, by resolution: $0.0375, $0.125, $0.1875 and $0.375. A 10-second clip at 720p is $1.25, against about $1.01 from Google's token math.

So on Sume you do not have a text meter to watch. The price you read in the catalog is the price you pay per second, and the Sume docs and catalog are the place to confirm it before a large run.