Voice so session transcripts can tell who's talking
Record about 45 seconds of yourself talking normally. The recorder uses it
to work out who said what in a session, so quotes end up attributed to the
right person. It's stored as a numeric voiceprint, not as audio.
Recording more than once genuinely helps. Each new sample is
averaged into your profile rather than replacing it — measured on real table
audio, one short sample wasn't enough to identify the speaker at all, and
four made it unambiguous. You get a different passage each time, so repeat
recordings cover different sounds.
A voiceprint matches best against the microphone it was made on, so one
recorded here on your phone or laptop and one recorded at the table are
both worth having — they're kept separately and the best match wins.
Read this — or say anything you like
0:00
Your browser won't allow microphone access here. It only permits it over
HTTPS — open the wiki by its https:// address and this will work.
Story Sections the heart of the page — reorder with the arrows
Link to any other page with [[slug]], or [[slug|shown text]]. Unknown slugs show as a dotted placeholder until that page exists.