Cortexa Voice

Cortexa Voice is the foundation layer of the Cortexa platform. It turns raw PTT and PMR radio audio into structured, searchable, and actionable intelligence. Every other Cortexa module — Safe, Ops, Signal — is built on top of Cortexa Voice.

What Cortexa Voice does

Ingests live audio from TETRA, DMR, P25, LTE-PTT, broadband-PTT, and analogue radio channels
Transcribes every transmission in real time, in the language being spoken
Detects language automatically per transmission — English, Arabic, Spanish, French, Dutch, German
Identifies speakers by callsign, team role, and acoustic signature
Threads fragmented transmissions into complete conversations
Recognises equipment IDs, location names, and site-specific vocabulary without manual configuration
Stores audio, transcript, speaker, channel, and timestamp as a linked evidence record

Why industrial radio transcription is different

General speech recognition products fail in industrial radio environments because they were not designed for them. Cortexa Voice was:

Handles compressed TETRA and DMR audio codecs (AMBE, ACELP)
Trained on industrial terminology: permit-to-work, lockout-tagout, hot work, confined space, equipment callsigns
Handles strong regional accents and mixed-language utterances within a single transmission
Manages high noise environments: machinery, vehicles, weather, compressed channels
Learns your site vocabulary automatically — equipment IDs, personnel callsigns, local procedure names

Output

Cortexa Voice produces a continuous structured stream of: speaker identity, channel, timestamp, language, transcript, confidence score, and linked audio. This stream feeds the Safe, Ops, and Signal modules and is available for search, export, and integration.

Evidence integrity

Every transcript entry is permanently linked to the audio that produced it. Supervisors can play back any transmission from any alert or report. Nothing is summarised or inferred without citation.