Signature Project

Talk2Me

A real-time AI voice conversation platform powered by Gemini Live API's bidirectional audio streaming. Features 7+ themed characters, outbound phone calls via Twilio, lead capture for insurance sales, and voice-controlled agent orchestration through tmux.

Gemini Live API FastAPI Twilio WebSocket Web Audio API PWA

Screenshots

Talk2Me Character Selection Hub
Character Selection Hub — 8 AI characters
Nicha Real Estate AI Advisor speaking
Nicha — Real estate AI advisor speaking
Chat Head integration on portfolio website
Embedded Chat Head — AI Nattapong on portfolio website

Impact

7+

Character Themes

Real-time

Audio Streaming

4

Industry Use Cases

PWA

Installable

Key Features

Bidirectional Voice AI

Real-time WebSocket streaming to Gemini Live API — user speaks, AI responds with synthesized voice instantly. No transcription latency, direct audio-to-audio.

7+ Character Themes

Matrix (Dali mask + digital rain), Call Center, Japan (kids game), Munich, Dhipaya Insurance (lead capture), Berlin (tmux tools), Interpreter, Real Estate. Each is a standalone PWA.

Outbound Calling via Twilio

Initiates real phone calls, bridges Twilio audio stream to Gemini Live API in real-time with mulaw-to-PCM conversion. AI talks to humans on the phone.

Insurance Lead Pipeline

PDPA-compliant consent management, phone validation (Thai 10-digit), auto-captures leads and sends to CRM API + Telegram notifications. Built for Dhipaya Insurance.

tmux Tool Calling

AI characters can read/write to tmux panes, enabling voice-controlled agent orchestration. Talk to your development team through voice.

Multi-Language Support

Thai and English with automatic language detection. Interpreter mode enables real-time translation between languages. Early adoption interest as AI interpreter solution.

More Projects