What is voice enrollment and do I need it?
Reading a short script so the app learns your voice and responds only to you. Not required to dictate, but it is what stops a livestream, a podcast or someone else in the room from driving your journal.
Voice mode listens whenever a field is focused, and a microphone cannot tell who is talking. Enrollment fixes that. You read about fifteen seconds of script, the app averages what it hears into a voiceprint, and from then on each piece of speech is scored against that print before anything is acted on. Speech that does not match is dropped.
Two things are worth knowing about how this is done, because they are unusual:
- It runs entirely in your browser. The speaker-recognition model is downloaded to your machine — about 100 MB the first time, cached afterwards — and the comparison happens locally. No audio is sent anywhere for this.
- Your recordings are not kept. What is saved is the mathematical fingerprint and its metadata: how many clips, how many seconds, when, and which model version. The audio itself is discarded.
If you trade alone in a quiet room you can skip it. If you stream, share a space, or leave a market feed playing out loud, it is the single most useful thing on this page.
