Output tokens in a 3 to 10 second Omni clip: 17,376 to 57,920
Gemini Omni Flash bills 5,792 output tokens per 720p second. A 3 s clip is 17,376 tokens and a 10 s clip 57,920. Dollars at Google and on Sume, per length.

A 3-second Gemini Omni Flash clip at 720p is 17,376 output tokens and a 10-second clip is 57,920, because Google's pricing page says billing is based on output tokens at 5,792 tokens per second of 720p video. At Google's $17.50 per million output tokens that is $0.30 for 3 seconds and $1.01 for 10 seconds. On Sume the same lengths cost $0.375 and $1.25 at 720p, because Sume bills a flat rate per output second instead of counting tokens.
Why the token view is useful
If you read Google's usage metadata, you see tokens, not seconds. Multiplying seconds by 5,792 lets you predict a bill before a job runs and check it after. The same multiplication is how you spot a clip that came back longer or shorter than you requested. Sume's 3 to 10 second window for Omni is the same range Google's docs give for the 3 to 10 second continuations when you extend a video, so the table below covers every length you can submit as one job on Sume.
The table
Each row multiplies the length by 5,792 tokens, then by $17.50 per million tokens for the Google column. The Sume column is the length times $0.125, which is Sume's 720p rate (the $0.10 provider list times 1.25).
| Clip length | Output tokens (5,792 per s) | Google cost at $17.50 per 1M | Sume cost at $0.125 per s |
|---|---|---|---|
| 3 s | 17,376 | $0.3041 | $0.375 |
| 4 s | 23,168 | $0.4054 | $0.50 |
| 5 s | 28,960 | $0.5068 | $0.625 |
| 6 s | 34,752 | $0.6082 | $0.75 |
| 7 s | 40,544 | $0.7095 | $0.875 |
| 8 s | 46,336 | $0.8109 | $1.00 |
| 9 s | 52,128 | $0.9122 | $1.125 |
| 10 s | 57,920 | $1.0136 | $1.25 |
What the table does not cover
The 5,792 figure is stated for 720p only. Google's Omni page lists 360p, 1080p and 4K outputs as well, and describes 1080p and 4K as upscaled, but its pricing page gives no token count for those, so this post does not extend the token column past 720p. For other resolutions use the Sume rates directly: $0.0375, $0.1875 and $0.375 per second at 360p, 1080p and 4K.
Google's pricing page also lists an input price of $1.50 per million tokens for text, image, video and audio input. Sume's docs describe its Omni billing as the provider list times 1.25 for each output second, so the billed quantity on Sume is output seconds.
Using the table for a budget
Suppose a campaign needs 40 clips of 6 seconds. Google's column gives $0.6082 a clip, so 40 clips is $24.33. Sume's column gives $0.75 a clip, so 40 clips is $30.00. The difference of $5.67 is the cost of a flat per-second rate, a single workspace balance, and a model id that stays callable after Google's preview ids end. If the campaign is only drafts, the 360p rate of $0.0375 a second brings the same 40 clips to $9.00.
A quick check in code terms
To confirm any row, take the length in seconds, multiply by 5,792, divide by 1,000,000 and multiply by 17.50. For 7 seconds that is 40,544 tokens, 0.040544 million, and $0.7095. Round at the end, not at each step, or a 100-clip batch drifts by a few cents.
On Sume, skip the token step: read usage.cost from the completed poll response and compare it with length x rate. If they differ, the job produced a different length than you asked for.
Sources
Related posts
More in Pricing
- Podcast sponsor ad reads: 36 spots of 450 characters, $1.08 on Sume
36 podcast ad reads of 450 characters cost $1.08 on Sume TTS at 3 cents each. MAI-Voice-2.1 lists $0.36 and Flash $0.24 for the same characters.
- What a 10-second video ad costs on each Sume model (Oct 2026)
A 10-second clip costs $0.125 on Grok Imagine Video 1.5 and $5.778 on Seedance 2.5 at 720p. All eleven model and tier prices for the 10 s minimum are below.
- Qwen-Audio 3.1 price cuts: rebuild your voice bill with three numbers
Qwen-Audio 3.1 cut prices about 70% for TTS, 85% for Realtime and up to 95% for ASR. How to recompute your bill, and where Sume TTS and STT sit.
- Can an API caller raise a Sume schedule's spend cap? No
A run request can lower generation_spend_cap_usd below the schedule cap but never above it. The six cases, including null and 0, in one table.
Written by Sume