Use cases · Meetings
Meeting transcription API: who said what, with timestamps
Speech Revolutions turns a meeting recording into a speaker-separated transcript with word-level timestamps, which is the input a notetaker, summary or action-item feature needs. It measured the lowest diarization error rate of five APIs on the AMI and NotSoFar meeting datasets.
from speechrevolutions import SpeechRevolutions
client = SpeechRevolutions()
# Returns as soon as the job is queued; the webhook says when it's done.
client.submit(
"standup-2026-10-21.m4a",
callback_url="https://you.example.com/webhooks/stt",
)Where it goes wrong
A meeting summary is only as good as its attribution. If the transcript gives the decision to the wrong person, the summary assigns the action item to the wrong person, and the user stops trusting the notes.
Meeting audio is hard: one laptop or room microphone, people at different distances, cross-talk, and calls that switch language when a colleague joins.
What the API gives you
- Speakers per word and per turn
- Each word carries a speaker, and utterances group them into turns, ready for an LLM prompt or a transcript view.
- Language detection
- Detected automatically, including when the meeting switches language part-way through.
- Webhooks
- Submit the recording with a callback_url and get a signed POST when it is done, instead of holding a connection open.
- Documents
- output_type="docx" or "pdf" returns a formatted transcript to hand to users.
- Data handling
- Audio and transcripts are deleted within 30 minutes of the job finishing, and customer audio is never used to train models.
Measured
On AMI-SDM, recorded business meetings on one distant microphone, Speech Revolutions measured 9.7% diarization error rate, against 25.3% for the next best API we ran. On NotSoFar, real far-field meetings, 10.9% against 23.3%.
Full results and method: speaker diarization measured across five APIs and all benchmarks.
Pricing
$0.003 per audio minute, $0.18 an hour, billed by the second. Speaker labels, word timestamps and every output format are included. New accounts get $10 of credit with no card required. Pricing details.