12 webinars of 47 minutes: the Sume STT line item is $5.64 a month

Sume lists speech-to-text at $0.01 per audio minute. Twelve 47-minute webinars are 564 minutes, so $5.64 of STT. What else is on the bill and how to avoid it.

5 min readSume
All posts

Transcribing twelve 47-minute webinars a month costs $5.64 in speech-to-text on Sume: 12 x 47 = 564 audio minutes at the public rate of $0.01 per audio minute. That is the transcription line only. If you run the transcript through a video inspect job, the compute reservation and platform fee are billed on top, which is why a clean audio file is the cheaper input.

The rate and what the docs say about it

The Video inspect page states the public rate for Sume STT 1.0 as $0.01 per audio minute and points to GET /v1/catalog for the live price. Without a duration hint, Sume reserves one minute, and the maximum duration hint is 600 seconds. The page does not describe how a partial minute is rounded, so for a 47 minute 20 second file read the live catalog before you promise a quote to a client.

STT arithmetic at $0.01 per audio minute (as of 2026-10-08)
VolumeAudio minutesSTT cost
1 webinar of 47 minutes47$0.47
12 webinars a month564$5.64
52 weeks of one webinar2,444$24.44
100 hours6,000$60.00

What a video-inspect transcript adds

The transcribe option on video inspect runs STT on the audio of a clip. Each inspect reserves its Modal compute ceiling, then captures its own container seconds times the Modal list times 1.25, plus the platform fee. The capture is never more than the hold, and the per-minute STT rate is added to that reservation. In other words, inspect is the right tool when you also want probe data or stills; it is not the cheapest way to get text out of audio you already have.

Source files can be at most 1,800 seconds (30 minutes) for inspect, so a 47-minute webinar does not fit in one inspect call at all. Cut it first or detach the audio in ranges; see the detach recipe in the related posts.

Make the transcript useful, not just cheap

Pass language_code (for example en or ko) when you know the language; otherwise STT auto-detects. Set segmentation.mode to sentence if you want sentence segments with no gaps in the shape of caption lines, and silence_split_seconds between 0.2 and 3 to control splits.

A silent clip fails with inspect_source_has_no_audio. Probe has_audio with frames set to false first; it is the cheap check.

Per month, per year

At 564 minutes a month the STT line is $67.68 a year. That is smaller than most people expect, and it is why the engineering effort should go to the inputs (clean mono audio, the right language hint) rather than to saving cents on the rate.

A monthly checklist

Before you run the month's batch: confirm each file has an audio track, cut intros and dead air, set language_code when the language is known, and split any file over the source limit. Then total the minutes and multiply by $0.01. Keep one row per file with its minutes and job id so you can reconcile the invoice.

If a webinar is 47 minutes 20 seconds, assume the partial minute may count as a full one until you verify the live rate. That makes 48 minutes, or $0.48, a one-cent difference per file.

Sources

Related posts

More in Pricing

All Pricing posts

Written by Sume