ProjectBuild an AI Voice Receptionist: the explanation · Module 13: Giving the receptionist a voice · part 15 of 23
Part 15 · Module 13: Giving the receptionist a voice
Module 13: Giving the receptionist a voice
23 min read·
About 35 minutes · Needs: Module 2 (the voice loop) and Module 10 (languages and normalisation) · Code: backend/app/voice/tts.py, backend/app/voice/audio.py, backend/app/voice/normalize.py
Where this fits
Text to speech (blue: this module) is the last stage of every turn; it relies on the language rules for Devanagari text.
Everything the receptionist decides has to reach the caller as sound. This module is the last stage of every turn: the agent's text goes in, 24 kHz audio comes out, one sentence at a time. It leans on the language work of Module 10, because the Hindi voice can only read Devanagari, and it hands its audio to the live WebSocket session of Module 14.
The rest of this project is yours to keep
This project is bought on its own, once, and stays readable afterwards, including the parts added to it later.