A voice-AI phone app. Pick a persona, drop your number, and the bot calls you for a real conversation — your voice is transcribed, run through a language model in character, and spoken back in its own voice, streamed both ways over a live phone line.
Really it's just an experiment with the voice model — I love working with voice and sound waves in general, and this is me playing in that space. It's also my attempt at building a real-time voice pipeline the hard way: stitching STT → LLM → TTS myself, no voice-to-voice modality model doing it all in one shot.
Twilio Media Streams · Whisper STT · Llama brain · Cartesia / Kokoro TTS · FastAPI + Next.js