Skip to content

Transcription requirements

Transcription is the first analytics stage: it converts a conversation's audio into the transcript that every downstream AI task depends on. Before configuring it, make sure the licensing, provider, network, and storage prerequisites below are met.

License requirements

Transcription requires a valid license for each user whose conversations are transcribed. The transcription entitlement is part of the Speech analytics license (target name: Conversation Analytics license; short code SPCH).

  • The license is assigned per user — see Tenant and user configuration.
  • The transcription job filters on this license, so conversations of unlicensed users are skipped.
  • Per-tenant caps on the number of assignable licenses and on transcription minutes are optional — also covered in Tenant and user configuration.

Speech-to-text providers

Transcription relies on an external speech-to-text service. The supported engine types are:

Engine Notes
MiaRec Speech API (v1 and v2) MiaRec's own transcription service. No language selection in MiaRec — language handling is decided by the service. Accepts files up to 250 MB.
Azure Speech-to-Text v3.0 Microsoft Azure. Large per-language picklist plus multi-language auto-detection candidates.
Google Speech-to-Text v1.p1.beta1 Google Cloud. Single-language selection per engine.

Regardless of the provider, the data flow is the same: audio recordings are uploaded to the transcription service, and the transcribed text is returned and stored in the MiaRec database.

Speaker separation depends on the audio: stereo recordings are transcribed per channel (agent and customer separated by design); for mono recordings, the Azure and Google engines offer a "recognize multiple speakers" diarization option.

Network requirements

The MiaRec server must be able to reach the transcription service over the internet (outbound HTTPS). Depending on your environment this may require firewall rules or proxy configuration. Each configured engine has a Test a Connection button to verify reachability — see System configuration.

Storage requirements

Transcripts are stored in the MiaRec database. Approximate storage requirement is 3–5 KB per minute of transcribed audio. Factor this into database sizing for high-volume deployments — transcript storage grows with recorded minutes, not with the number of conversations.

Next steps