37-second AI avatar video: $6.81 standard, $9.07 plus, $20.35 max

A 37-second talking avatar video on Sume costs $6.81 at the standard tier, $9.06 at plus (rounded up from $9.065) and $20.35 at max, using the per-second prices in the public catalog as of 2026-10-09. Thirty-seven seconds is a typical length for a product explainer with one claim, one proof point and a call to action, and it sits inside the 4-60 second window the docs set for avatar videos.

The three tiers share one request shape. Only the quality field changes, so you can price the same script three ways before you commit.

The arithmetic

Avatar video is billed per output second. The catalog lists standard at $0.184 per second, plus at $0.245 and max at $0.55. Multiply by 37 for the totals below. Sume rounds reservations up to whole cents, so the plus figure shows as $9.07 even though the raw product is $9.065.

The default is plus, not standard

The docs say quality defaults to plus when you omit it. If you budgeted with the standard rate and left the field out, the reservation is $9.07, not $6.81: a gap of $2.26 on one clip. Across a batch of 20 such videos that is $45.20. Set the field explicitly in every request so the number in your spreadsheet is the number on the invoice.

The docs describe standard as the fastest Sume execution path and max as the highest quality tier with slower turnaround. Plus is the balanced path.

Staying inside the window

Sume accepts scripts and multi-scene plans when it estimates the target duration at 4-60 seconds inclusive. A 37-second estimate passes. A script it estimates at 61 seconds does not, and the docs tell you to shorten it or split it into several jobs. The estimate comes from the script, so a script you believe runs 37 seconds aloud can land a few seconds either side. Leave headroom near the limit.

If the look matters before the spend, use an avatar video preview. It produces first-frame stills before any full render, and you pick the final tier when you call generate-video.

• Set quality explicitly: standard, plus or max.
• Keep scripts under the 60-second estimate; split longer ones into jobs.
• Poll GET /v1/jobs/:id/status, then fetch /result when it completes.

When 37 seconds is the wrong shape

If you only need a mouth to follow audio you already have, a lip-sync or talking-still route can cost less per second. See which Sume avatar route fits for the choice between face swap, motion control and lip sync. The figures above are for a fully script-driven video, with avatar, voice and scene handled in one job.

A planning shortcut

For quick estimates, use whole-dollar anchors per 10 seconds: about $1.84 at standard, $2.45 at plus and $5.50 at max. Add them up for the length you need. Thirty-seven seconds is three and seven tenths of those blocks, which is how the totals above come out. When a client asks for a price before a script exists, quote the plus figure and state that standard is cheaper and max is higher, so no one is surprised by the default.