Transcribe 1,000 four-minute voice memos on Sume STT: $40.00

1,000 voice memos of 4 minutes each is 4,000 audio minutes. At Sume's public speech-transcription rate of $0.01 per audio minute that is $40.00, or $0.040 per memo. Send the memo length as duration_seconds so the reserve matches the clip.

The arithmetic at four batch sizes

The rate card lists speech transcription at $0.01 per audio minute. The video inspect docs describe the transcript option on that route: without duration_seconds Sume reserves 1 minute, and the maximum hint is 600 seconds. A 4-minute memo is 240 seconds, so it fits under that hint.

The table scales the same arithmetic across batch sizes. Memo length is held at 4 minutes.

What the number leaves out

Two things the rate does not cover. It does not include the compute ceiling the inspect route reserves for the clip, so budget a little above the pure per-minute figure when you go through that route; Sume captures actual cost and never more than the hold. And it is priced per audio minute of the source, so silence and hold music count.

Sume bills model and tool usage at the published rate on top of the plan fee, so the plan price is separate from every number in this post.

Worked example and checks

A budgeting habit that helps here: price the batch before submitting it, then compare the captured total with the estimate. Because Sume reserves at submit and captures actual cost, the reserved figure is the ceiling you can lose track of, and the captured figure is what you pay. Log both per memo and the difference tells you whether your duration hints are accurate.

The same arithmetic works for any speech source. Divide total audio minutes by 60 for hours: 4,000 minutes is 66.7 hours, so the rate works out to $0.60 per hour of audio, which is the number to put next to other vendors' hourly rates when you compare. Each vendor's own page sets its own unit, so convert before comparing.

How to batch it

Chunk by memo, not by hour: one job per memo keeps each under the 600-second hint, makes a failed memo a single retry, and gives you a per-file cost line. Use language_code when you know the language, since otherwise STT auto-detects.