Report
Phonon-2 introduced as a 164MB speech recognition model
The announcement claims Phonon-2 is more accurate on average than OpenAI’s Whisper large, a model 10 times its size.
TLDR
A September 29 post introduces Phonon-2 as a 164MB speech recognition model with open weights under CC-BY-4.0. It claims better average accuracy than OpenAI’s Whisper large, which it describes as 10 times bigger, and says Phonon-2 can transcribe an hour of audio in 20 seconds on a MacBook Air.
Combined views
—
1 Source, first seen ago
