How is a sume/auto video job billed if Auto has no price?
sume/auto is not a priced model: billing keys off the family that served the job, at list x 1.25, reserved on submit. What the receipt shows and what it hides.

sume/auto is a routing name, not a model, and it has no price of its own. Sume's internal video docs state that billing keys off the resolved family, so the amount you pay is the per-second rate of the family that ran, at the provider list times the 1.25 house margin that every Video Router row carries. The submit and poll responses echo sume/auto verbatim, and the family that served the job is not disclosed.
That raises a fair budgeting question: if you cannot see the family, how do you know the price? You can, because Auto's envelope is narrow and documented.
What does an Auto clip cost in practice?
Sume's video generation docs say Auto defaults to Gemini Omni Flash 1.1, with 3 to 10 seconds, 360p to 4K, and 16:9 or 9:16. Gemini Omni Flash 1.1 bills per output second by resolution, so you can price an Auto request from its resolution and length. The table shows an 8-second clip, which is the Auto default length, at each resolution.
| Resolution | Per second | 8 seconds |
|---|---|---|
| 360p | $0.0375 | $0.30 |
| 720p | $0.125 | $1.00 |
| 1080p | $0.1875 | $1.50 |
| 4K | $0.375 | $3.00 |
Where do I read the real amount?
The poll response carries usage.cost, the Sume billable amount, and the same job is readable at GET /v1/jobs/{id}/result in Sume's normal envelope. Funds are reserved at submit and the final amount is captured when the job completes, so the number to trust is the one on the finished job, not the estimate. See Jobs and results for the lifecycle.
Send an Idempotency-Key and a retry returns the original job instead of creating and billing a second one.
When should I stop using Auto for cost control?
Auto is the right default when you want a working 3 to 10 second clip with sound. It is the wrong tool when the price is the decision: Wan 3.0 at 480p is $0.0625 per second, below Gemini Omni Flash 1.1 at any resolution except 360p, and it runs longer clips. Pin the row you priced, and keep Auto for the requests where the envelope already fits.
Sources
Related posts
More in Pricing
- How long can a Reel be? 3 minutes, and what each render costs
A Reel can run up to 3 minutes as reported in October 2026. A Sume Timeline render of that length costs $0.30, billed at $0.10 per output minute rounded up.
- How many minutes of AI avatar video does $100 buy on Sume?
At Avatar 1.0 per-second rates, $100 buys 9.06 minutes at standard, 6.80 at plus and 3.03 at max. In 30 and 60 second clips, 18 to 2 whole videos.
- How many seconds of AI video does $20 buy on Sume, by model?
At Sume's billable per-second rates, $20 buys 320 s of Wan 3.0 at 480p but 53 s of Gemini Omni Flash 1.1 at 4K. Full table by model and resolution.
- How much to top up a Sume wallet: what $10 and $1,000 buy
Sume wallet purchases run $10 to $1,000. What each buys in avatar seconds, music, cutouts and captions, and a sizing rule: burn plus the largest hold.
Written by Sume