Key Info

A developer explored using Jev as a semantic VAD (voice activity detection) for real-time voice AI, integrating it with Agora ConvoAI to determine whether a user has actually finished speaking.

Highlights

  • Uses live transcript and recent conversation as context for Jev's decision
  • Plugged into Agora ConvoAI (@AgoraIO) for real-time voice AI
  • Aims to solve the problem of AI bots that don't listen well or interrupt
  • Described as handling this task surprisingly well