1 paper · 1 filter
Ashish Thapa
This paper describes an audited neural text-to-speech stack that runs from phoneme IDs to 22.05-kHz PCM on general-purpose microcontrollers. Its deployed graph has 567,008 paramete…