I work
with voice.从声音出发，
看世界。
Researcher and builder exploring the space between speech, music, physiology, and AI.
The room
is yours.你的客厅，
就是 KTV。
Tired of paying for KTV subscriptions and reaching for a VPN just to sing? Now you can host your own KTV.
KTV does not provide or distribute music videos. You supply the media you own or are licensed to use and are responsible for its rights.
Barely
recorded.
Played again.录音寥寥，
古琴之声，
再次响起。
The guqin was barely recorded. I gathered nearly 100 hours of it to train an autoregressive music model that never stops playing. Every piece in the listening room is AI-generated.
Hear a voice.
See it move.听见人声，
看见声带在动。
Every sound starts inside the body. WaveEGG listens to a voice recording and recovers the physiological signal behind it, with no sensors on the body.
Every voice
has a shape.每一种声音，
都有自己的形状。
Voice Mapping turns specialist measures into an approachable portrait of vocal qualities. Compare illustrative voice profiles and learn what the dimensions mean.
The rest of
the story.故事，
还在继续。
Research papers, things I have built, and what I do when I leave the lab.