Skip to main content

Setting Up Voice Transcription

This guide covers microphone setup, consent requirements, and starting real-time voice transcription.

Microphone Setup​

Follow these recommendations for accurate transcription:

  • Use an external microphone rather than your laptop's built-in microphone
  • Position the microphone between you and the client, or use a lapel microphone
  • Reduce background noise and test your microphone volume before the session begins
TypeBest ForNotes
USB condenserIndividual sessions in an officeGood balance of quality and cost
Boundary microphonePlaced on a table between speakersWorks well for couples sessions
Lapel microphoneTelehealth sessionsMinimises background noise

Browser Permissions​

When you first use voice transcription, your browser will ask for microphone access:

  1. Click Allow when prompted for microphone permission
  2. If you previously denied permission, go to your browser's site settings for Praxi and enable microphone access
  3. Praxi checks your audio input automatically and displays an audio quality indicator
Checking Audio Before a Session Open a session and click the microphone

icon in the toolbar. The audio quality indicator shows your input level, clipping warnings, and background noise detection. Adjust your setup until the indicator shows a green status. :::

Recording and transcribing a clinical session requires explicit client consent under the Privacy Act 1988:

  1. Before starting transcription, click Start Recording in the session interface
  2. Praxi checks whether the client has an active transcription consent on file
  3. If no consent exists, a consent prompt is displayed -- the client can acknowledge consent digitally
  4. Consent is logged with a timestamp against the client record and is valid for 12 months by default
  5. You can request consent from the client in advance by sending a consent request notification from their profile -- the client can then grant consent before the session

A client can revoke transcription consent at any time. If consent is revoked, transcription is no longer available for that client's sessions until new consent is granted. You can revoke consent on the client's behalf from their profile.

Separate AI Consent Transcription consent and AI processing consent are

separate. A client may consent to transcription (speech-to-text) but decline AI analysis (note generation, live guidance). If a client consents to transcription but not AI processing, the transcript is captured but AI features such as note generation and live clinical guidance are not available for that session. :::

Consent Renewal Transcription consent is valid for 12 months. When a

client's consent is nearing expiry, you can renew it from their profile or from the consent prompt that appears when starting a recording. :::

Starting Transcription​

  1. Open the session from the calendar or client profile
  2. Click Start Recording in the recording card
  3. Confirm or obtain client consent when prompted
  4. Praxi checks your audio quality and displays an indicator for microphone level, clipping, and connection status
  5. Transcription begins -- you will see a live indicator showing that audio is being captured
  6. Interim (partial) transcription segments appear in the transcript panel in real time
  7. Final (stabilised) segments replace the interim text as the system confirms accuracy

To stop transcription, click the stop button or end the session.

Speaker Identification​

Praxi uses speaker diarisation to distinguish between different speakers in the session:

  • The system automatically detects speaker changes
  • Label each speaker (e.g., "Practitioner", "Client") when prompted after the first few segments
  • You can swap speaker labels after recording if the system attributed a segment incorrectly, including bulk swapping for multiple segments at once
  • Speaker labels are used by the AI note generation to attribute content correctly

Multi-Speaker Sessions (Couples and Groups)​

For couples or group sessions, Praxi supports multichannel transcription where each audio channel is mapped to a specific speaker. This eliminates diarisation errors by physically separating audio streams. Each participant requires their own microphone for best results.

Multi-speaker transcription requires a PRO or ENTERPRISE subscription. You can also enrol speaker voiceprints for automatic speaker identification in future sessions (requires explicit consent as voiceprints are biometric data).

Audio and Transcript Retention​

Praxi follows strict data minimisation practices for session recordings:

  • Audio files are not kept by default: audio is destroyed as soon as the transcript is ready. A practice can choose to keep audio for up to a set number of hours (up to 720) so practitioners can play back moments while verifying the note. Kept audio is destroyed when the note is signed, or when the window ends, whichever comes first
  • Transcript text is redacted 7 days after the clinical note is signed -- only metadata (duration, segment count, speaker statistics, word count) is retained. Verbatim text content is permanently deleted
  • Clinical notes are retained for 7 years in line with AHPRA record-keeping requirements
  • Session memories are retained indefinitely as they contain structured summaries rather than verbatim content

Audio cleanup runs hourly and transcript redaction runs daily.

Data Minimisation Audio recordings and transcript text are temporary

working data. Once the clinical note has been reviewed and signed, the audio and transcript are no longer needed and are automatically cleaned up. :::

Accuracy Tips​

  1. Speak clearly and avoid speaking too quickly
  2. Minimise crosstalk -- try not to speak over each other
  3. Reduce background noise -- close doors, silence phones
  4. Use Australian English -- the transcription engine is configured for en-AU

Was this page helpful?