Chimera Vocal Synth
A voice grown from a simulated throat — sing, speak, and beatbox in your browser.
Project intro
Chimera is a physical-modeling voice synthesizer that runs in your browser. Every sound comes from a simulated airway — lungs, vocal folds or a syrinx, a branched vocal tract, lips, and nose — not from samples and not from a neural model. The engine is written in Rust, compiled to WebAssembly, and runs in an AudioWorklet; the same code runs natively as a command-line tool for offline renders.
Play it by dragging the tongue, lips, and jaw in a live cutaway of the head, with a vowel pad, a computer keyboard, or Web MIDI. Speak English text or IPA and it turns the words into gestures and prosody. Sing by typing lyrics and letting it fit each line to a tune picked by its mood, then record a 9:16 video clip to share. Beatbox on pads and a step sequencer where every drum is a mouth gesture, or open the Bestiary and hear a songbird, a seal, an alligator, and an elephant. Lyrics work in English, Korean, and Russian, and male, female, and child bodies blend continuously.
More than 2,500 developers starred it in its first eleven days, with 318 forks and 219 commits, and the demo at chimera.audio runs the real app in the browser. It is MIT licensed, and its own docs score the voice against real singers and speakers, phone by phone, so you can hear exactly how close the simulation gets.
It is an instrument, not a content pipeline: there is no video editor, no captions, and no publishing step — the app records a shareable video clip and stops there. Clean, controllable speech is the hard part — the simulation is frank about the gap — and the AudioWorklet needs a secure context, so you open it on localhost or over HTTPS.
Actions
Tech stack
Project info
- Project category
- Music & Audio
- Project status
- Online
- Collaboration intent
- None