Clap once and the piece learns your room. Hold one note, then hear it inside a stone dome, at the focus of a planetarium, aboard the Space Station, in a habitat under the Moon's surface, outside the airlock, and on Mars.
The question
Acoustics is usually taught as equations and the solar system as pictures. What if you could hear the difference between places, in your own voice, in under two minutes? And could the strangest facts (that Mars has two speeds of sound, that the Moon has none) land as feelings instead of trivia?
What we made
A five-room journey built on one recording of your voice:
| Room | What you hear | How it's made |
|---|---|---|
| Stone dome | Your note hanging for seconds after you stop | A measured impulse response of Hamilton Mausoleum (OpenAIR library, University of York), convolved with your voice |
| Planetarium | Every echo returning to you as one sharp slap | A hard 15 m plaster dome simulated with ray tracing and image sources, you at the center. Modern domes are perforated to kill this; the piece says so |
| Space Station | A close, bright, boxy room with fans that never stop | A simulated padded cylinder the size of the Destiny lab, plus a synthesized fan bed |
| Habitat, then the airlock | Soft walls, then nothing | A simulated inflatable, then a gain of zero. No medium, no wave |
| Mars | Quieter, duller, and the high and low parts of your voice arriving at different times | A filter model fitted to Perseverance's microphone data: −20 dB, a 240 Hz crossover with a speed-of-sound delay difference, and distance-scaled absorption |
The sound is drawn as light: rays leave the singer, bounce off the room, and fade with its decay. A circle returns every ray to its center, which is also why the planetarium slaps. On Mars the high-frequency rays are blue, travel slightly faster, and die sooner, because they do.
At the end you get a postcard: one note through five rooms, as audio and as an image where each room's decay is drawn on a shared 12-second clock.
What's true, and what's labeled
- The dome is a real measurement. The reverberation time is about 12 s at 500 Hz and 9 s at 1 kHz.
- The simulated rooms are built from stated dimensions and materials with pyroomacoustics, and are labeled “simulated” on screen.
- Mars numbers come from Maurice et al., Nature 605 (2022): about 240 m/s below the CO₂ relaxation frequency (about 240 Hz) and 246–257 m/s above 2 kHz; sounds about 20 dB weaker than on Earth; at 8 kHz, −40 dB by 8 m where Earth loses −20 dB. The model fits those points; the band between 240 Hz and 2 kHz is interpolated.
- The pictures are illustrations, not photographs of the measured spaces.
What happened in testing
The first automated run recorded three seconds of silence, and the normalizer dutifully amplified the noise floor. The piece now refuses a silent take and asks you to sing a little louder. The first postcard clipped; a limiter and a normalization pass fixed it. Measured spectra of the renders confirmed the Mars roll-off (−16 dB at 8–12 kHz relative to the habitat at 4 m) before anyone listened.
What's next
Measured whispering-gallery and dome positions from St Paul's Cathedral are available in the same library and could replace the simulated planetarium. A recorded stereo pair would let the postcard carry the room's width.
Tools
Web Audio API (convolution, biquad crossovers, delays, an AudioWorklet recorder, offline rendering), Canvas, Vite + TypeScript, pyroomacoustics for the simulated rooms, Kokoro TTS for the guide's voice, Playwright with a fake microphone for end-to-end tests. Frames and voice were generated on an NVIDIA DGX Spark; nothing in the piece calls an external API.