Echo Cancellation During Transcription

This is not essential reading for therapists. You do not need to understand acoustic echo cancellation to use AirNote. But if you are curious about what happens between a remote therapy session and its transcript, it is one of the quieter technical details worth knowing about.

Acoustic echo cancellation, usually shortened to AEC, is audio processing that removes sound from a computer’s speakers when that same sound is picked up again by its microphone. For online therapy, it can make a session recording cleaner and the transcript easier to follow.

What is acoustic echo cancellation?

During a remote therapy session, a client’s voice may come through your Mac’s speakers. The Mac’s microphone captures your voice, but it can also hear the client’s voice bouncing back from the speakers a fraction of a second later. The recording may therefore contain the direct client audio and a delayed, room-coloured copy of it.

AEC compares the audio sent to the speakers with the sound arriving at the microphone. It continuously estimates what that reflected version should sound like, then reduces the matching echo while preserving new sound, including the therapist’s speech. It is closer to recognising a known reflection than simply turning the microphone down.

Why echo affects remote therapy transcription

If speaker echo remains in a remote therapy recording, phrases may appear twice, words can overlap, and it can become harder to tell who said what. This is particularly unhelpful in a clinical conversation, where small differences in wording and speaker attribution can matter.

An AI transcription or speech-to-text system can only work with the audio it receives. Cleaner separation between the therapist and client gives transcription and speaker labelling a more reliable source. That can reduce avoidable errors and give any subsequent clinical note generation a clearer transcript to work from.

AEC does not fix every audio problem. It cannot repair a poor internet connection, a very quiet microphone or unrelated background noise. Headphones still provide the most reliable physical separation between speaker and microphone. AEC is most valuable when the Mac’s built-in speakers are being used.

How AirNote uses AEC

For an eligible remote session using the Mac’s built-in speakers, AirNote records the microphone and system audio separately. It uses WebRTC Acoustic Echo Canceller 3 (AEC3) to compare the two in very short sections of audio. The cleaned microphone track then becomes the source for live and final transcription.

AirNote checks the actual audio route before using this processing. It is not applied to in-person or microphone-only recordings, and headphones, external speakers or an uncertain output route use the original microphone track. This keeps the feature focused on the situation it is designed to improve.

Private, on-device echo cancellation

The echo-cancellation step is fully private and runs directly inside AirNote on the user’s Mac. The microphone audio and speaker reference do not need to be uploaded to a cloud echo-cancellation service. For therapy audio that may contain protected health information (PHI) or other sensitive clinical information, keeping this processing on the device removes an unnecessary transfer of session audio.

This follows the same privacy-first principle as AirNote’s local AI transcription: use the Mac’s own processing power where it can do the job well, rather than sending raw therapy recordings elsewhere.

Built to improve audio, not risk the recording

AirNote keeps the original microphone and system recordings independently of the cleaned audio. If echo cancellation cannot complete safely, AirNote falls back to the original microphone track rather than relying on a partial enhanced version. Recording can continue normally, because AEC is an improvement rather than a single point of failure.

Most therapists and psychologists should never need to think about acoustic echo cancellation. Its value is deliberately practical: fewer repeated voices, clearer speaker separation and better input for private therapy transcription, while the processing itself stays on the Mac. It is a small part of AirNote, but an important one for making remote session transcription more dependable.