Skip to main content

Introducing AI Voice Transcription & Live Clinical Guidance

· 4 min read
Praxi Team
Praxi Team
Product & Engineering

We are excited to announce the launch of AI-powered voice transcription in Praxi, bringing real-time clinical documentation assistance directly into your sessions.

Real-Time Streaming Transcription​

Praxi now captures speech during clinical sessions and converts it to text in real time using Australian English language models with speaker diarisation. The streaming pipeline processes audio in 100-millisecond chunks, delivering interim and final transcript segments to your screen with minimal latency.

Transcripts appear in a dedicated drawer alongside your session interface, clearly distinguishing between practitioner and client speech. You can review, edit, and annotate transcript segments at any point during or after the session.

Live Clinical Note Generation​

As the session progresses, Praxi generates structured clinical note sections progressively — approximately every five minutes. These draft sections follow your preferred note template (DAP, SOAP, or custom formats) and are presented for your review. Nothing is finalised without your explicit approval.

The structured extraction pipeline identifies nine categories of clinical fact from the transcript, including presenting issues, interventions used, client responses, risk indicators, homework assigned, and session themes.

Clinical Guidance from Praxi​

During sessions, Praxi can surface real-time clinical suggestions based on the conversation. These include evidence-based intervention prompts, risk alerts when concerning language is detected, and theme tracking across sessions. You can also ask Praxi questions directly using the "Ask Praxi" feature in the guidance panel.

All AI suggestions are clearly marked with confidence indicators so you can assess their reliability at a glance.

Session Memory​

When you sign a clinical note, Praxi automatically consolidates key information into session memory. This memory is available to Praxi in future sessions, providing longitudinal context about the client's therapeutic journey — including themes, progress, homework completion, and treatment milestones.

Theme tracking follows a lifecycle from emerging to active to resolving to resolved, giving you a clear view of the client's trajectory over time.

Voice transcription in Praxi is built with Australian Privacy Act 1988 compliance at its core:

  • Recording consent is mandatory — Transcription cannot begin until the client has provided explicit consent, recorded in their client record.
  • Separate AI processing consent — An additional consent flag is required before AI analysis (note generation, theme tracking, clinical guidance) is applied to transcripts.
  • Configurable audio retention — Audio retention is a per-organisation policy: from a 72-hour default down to a zero-retention mode that destroys audio the moment transcription completes. After the configured retention expires, audio is permanently deleted.
  • Transcript retention — Transcript retention is configurable per organisation, and transcripts are encrypted at rest. Once the clinical note is signed, the transcript is redacted and only the structured note remains.
  • Couples and group sessions — Multi-speaker sessions require consent from all participants. Three consent models are available: no secrets (all content shared), secrets (practitioner-only segments), and case-by-case (per-session decision).

Your practice's consent settings are managed in Settings > AI > Consent.

Audio Quality Tips​

For the best transcription accuracy:

  1. Use a quality microphone — Built-in laptop microphones work but external USB microphones or headsets significantly improve accuracy.
  2. Reduce background noise — Close windows, turn off fans, and use a quiet room.
  3. Speak naturally — The Australian English model handles natural speech patterns well. No need to speak slowly or artificially clearly.
  4. Monitor the quality indicator — The audio quality badge in the session toolbar shows real-time mic level and connection quality. Green means good; amber suggests adjustments may help.

Availability​

Voice transcription is available as a per-seat add-on on all paid plans — see the pricing page for current add-on tiers and inclusions. Batch (post-session) and streaming (live notes) modes, dictation, and audio upload are all included; multi-speaker couples/group transcription follows your plan's couples/group feature availability.

To get started, navigate to Settings > AI > Transcription to enable AI processing consent, then start a session and click the microphone icon in the session toolbar.

Learn More​