68×
faster than real time
measured Aug 24, 2026
9.8%
diarization error rate
vs. human-labeled ground truth
$0.10
per hour of audio
at any volume
99
languages
same pipeline, same price
Compare the full bill
Elsewhere, knowing who spoke costs extra. Here it's in the price.
The real cost of 1 hour of audio with speaker identification, at each provider's public pricing.
Public prices as of Aug 24, 2026. Groq charges $0.04 per hour but with no speaker identification at all — transparency goes both ways.
Every word in the transcript. Every voice with a name. One flat price.
How it works
Three steps, no surprises
01
Create the job
Send your audio URL or request a signed upload URL. Up to 600 MB and 4 hours per file.
POST /v1/jobs
{ "audio_url": "…",
"webhook_url": "…" }02
We process it
Async queue on GPUs. A 3-hour episode is done in minutes, not hours.
→ your webhook
{ "id": "…", "status": "done" }03
Fetch the result
JSON with speakers and timestamps, or ready-made captions. Pick the format in the URL.
GET /v1/jobs/{id}/transcript
?format=srt | vtt | txt | jsonPricing
One line. This is it.
$0.10
per hour of audio
Billed per second processed, prorated. Failed jobs are never charged.
Start with 2 free hours- Transcription and diarization, always together
- JSON, SRT, VTT, and plain text at the same price
- Webhooks with retries for up to 21 hours
- Prepaid balance, measured in hours of audio
- Credits never expire, no subscription
No tiers, no sales calls: the price is the same from your first hour to your thousandth.
FAQ
What you'd want to know before integrating
Get started
Paste the curl, create an account, transcribe 2 hours on us.
2 free hours · no credit card · credits never expire