Voice platform
pyannoteAI
Speaker diarization and identification API for knowing who said what in calls and meetings.
- Pricing at a glance
- Developer €19/mo, Starter €99/mo with usage credit; Enterprise custom
30-day free trial with no credit card, including API and playground access.
Know who said what before you analyze calls
What you get
Add diarization, speaker identification and speaker-attributed transcripts to call analytics or QA, via API or on your own infrastructure.
What to account for
It is an analytics component, not an agent. Voiceprints raise biometric consent questions, and accuracy claims are vendor-reported.
What the platform provides
- Speaker intelligence
- Streaming and batch diarization, voiceprint-based speaker identification and speaker separation.
- Pipeline role
- STT orchestration that merges diarization with your chosen transcription provider; an analytics component rather than an agent.
Connections & compatibility
- Connects with
- pyannote.audio, Any STT provider, On-premises deployment
Pricing & terms
Developer is €19 a month and Starter €99 a month, each including the same amount in usage credit; the 30-day free trial covers up to 170 hours of audio. Enterprise pricing is volume-based with on-premises options. Per-hour rates beyond the included credit were not shown on the pricing page.
Before committing
Run your own recordings with crosstalk and short turns, and compare the error against the open-source model you may already use.
- What diarization error do you see on our call recordings compared with our current setup?
- How long are audio and voiceprints kept, and how do we delete a speaker's voiceprint?
- Which STT providers does orchestration support, and how are timestamps reconciled?
- What does usage beyond the included credit cost per hour of audio?
Controls to confirm
pyannoteAI describes itself as GDPR-native with on-premises deployment available. Confirm audio retention, whether voiceprints are stored and how they are deleted, and your legal basis for biometric speaker identification before enrolling customers or employees. Morak has not audited these claims.
What to consider next
A different approach
Consider AssemblyAI if you want transcription and speaker labels from one speech API.
A different approach
Investigate Deepgram for fast transcription with built-in diarization.