Transcription

Exovo transcribes speech on your own hardware — the Whisper speech-recognition model runs inside the system, so recordings and voicemails become text without any audio ever leaving your server. No cloud transcription service, no per-minute fees, no third party hearing your calls.

Enabling transcription

Admin → Integrations → Transcription: enable the engine and download the language model (a one-time download of roughly half a gigabyte; budget ~2 GB of RAM for the engine while it's transcribing). Language is configurable; the model handles the major European languages well.

What gets transcribed

• Voicemail — messages are transcribed automatically, so the text arrives alongside the audio in voicemail-to-email and the web client.
• Call recordings — select one or more recordings in the recordings library and choose Transcribe to queue them; the text appears against each recording once the background worker finishes. The CRM integration also transcribes automatically and writes the transcript into the contact's call record on CRMs that support it.

Practical expectations

Transcription runs in the background on the box's CPU — a busy call center generates transcripts minutes behind live, not instantly. Accuracy is genuinely good on clear business audio and degrades with speakerphone echo and crosstalk, the same way a human transcriber would. Treat transcripts as searchable evidence of what was said, not verbatim legal records.