Nectarine

mirrorbird
(18:22:58)
sometimes comes up with amusing mistakes for place names and so on
mirrorbird
(18:22:47)
it is generally very good. gets about 97% correct -- no voice training required
faraday
(18:22:37)
like the old days of the telegram [STOP]
mirrorbird
(18:21:56)
and weirdly, every time, it ends the transcription with exactly two: [BLANK_AUDIO] [BLANK_AUDIO]
faraday
(18:21:52)
[whispers continue]
mirrorbird
(18:21:48)
i've started using this AI speech-to-text thing, for dictation
faraday
(18:21:42)
[whispers]
faraday
(18:21:34)
[dog barking]
faraday
(18:21:16)
[roars] [birds chirping] [leaves crackling] etc
faraday
(18:20:13)
so now I know plenty of words like "