Audio Streaming - Using WebRTC for building your own Voice AIs
summary
Shows a working prototype that streams browser mic audio to a Node back end, converts 44.1kHz→16kHz, sends chunks over sockets, runs Google Dialogflow and Cloud Speech-to-Text, then synthesizes with Cloud Text-to-Speech and plays via AudioBuffer. Covers RecordRTC, socket.io-stream, WebRTC over HTTPS, handling iOS autoplay quirks, temporary WAV files, mono channels, and a knowledge-base FAQ import. Live demo handles multilingual input and sustained streams far longer than smart‑speaker limits.
AI-written summary · about
chapters (7)
- 0:00Start
- 1:11Google Assistant facts
- 2:55Use Cases for custom Al voice assistants
- 4:58The idea: Airport Self Service Kiosk Demo
- 8:05Architecture
- 11:35Google Cloud
- 18:31open sourced my code
related talks
26m#javascript Audio/Video Streaming: #WebRTC APIs 101 explained
43mNow Hear This! Putting Real-Time Voice, Video and Text into Rails
23mAudio and Video Manipulation
25mPerforming Transcription of WebRTC Media Streams in Browser
38mReal-time Communication on Mobile
31mWebRTC: Real-time communication without plugins
24mMaking WebRTC awesome
19mMixing voice & data