11-minute clip, 10-minute transcript cap: two passes cost $0.15

A transcript on Sume covers at most 10 audio minutes per call, so an 11-minute webinar clip needs two passes. Trim it into a 10-minute part and a 1-minute part at $0.02 each, then transcribe 10 minutes and 1 minute at $0.01 per minute. The Sume-priced total is $0.04 + $0.11 = $0.15, plus the compute that each inspect call reserves and settles on its own.

Where the 10-minute limit comes from

The Video inspect guide says transcribe: true runs Sume STT 1.0 on the audio of the clip. The public rate is $0.01 per audio minute, a request without duration_seconds reserves 1 minute, and the maximum duration hint is 600 seconds. An 11-minute clip is 660 seconds, so a single hint cannot describe it.

The Video trim guide prices a range cut of one hosted clip at $0.02 per job. That is a flat price, so cutting at 600 seconds costs the same as cutting at 60.

The arithmetic

Split the clip at 600 seconds. Part A is 0 to 600 seconds and part B is 600 to 660 seconds. Both trims produce new Sume-hosted MP4 files, which the inspect route needs as input.

What the $0.15 leaves out

Each inspect call also reserves a Modal compute ceiling and settles at its own container seconds, with the platform fee. The docs say the capture is never more than the hold. That compute is not a fixed number in the docs, so the table above covers only the quoted rates. Read the debited amount from the usage ledger after a first run.

• Without transcribe: true, the inspect call has no transcript charge.
• If you send language_code, segmentation or duration_seconds without transcribe: true, the API returns 400 video_inspect_transcribe_required.
• A silent clip returns inspect_source_has_no_audio; check probe.has_audio first.

Scaling it up

Ten 11-minute webinars cost 10 x $0.15 = $1.50 at the quoted rates. A 25-minute recording needs three parts (10, 10 and 5 minutes): three trims at $0.02 plus 25 transcript minutes at $0.01 is $0.06 + $0.25 = $0.31. Because the trim price is flat per job, you pay for the number of parts, not the length of each.

Checking the split before you pay

Before paying for transcripts, confirm the clip length and whether it has audio. A probe-only inspect tells you probe.has_audio and the duration, so you know whether the clip is 660 seconds or 700. The number of parts is the ceiling of the length divided by 600, so an 11 minute clip is two parts, a 20 minute clip two parts, and a 20 minute and one second clip three parts.

The table below shows how the part count and cost move for common webinar lengths at the quoted rates.