Eight NASA datasets, pressed onto cassettes. Insert a tape and the planet plays back as sound.
At least 2.2 billion people live with a near or distance vision impairment, and about 43 million are blind. A temperature curve or a sea ice map usually reaches people through the eye alone.
Sonification maps the numbers to pitch, rhythm and texture. Anything we can show as an image we can also play, so the same dataset can be read by ear on an ordinary laptop with a keyboard.
Choosing an era, loading a tape and hearing sea ice turn into a chord stays with an audience far longer than a static figure on a slide. That is why the data is a deck and not a dashboard.
VOXTERRA is a project by ANKATECH. Figures: WHO fact sheet on blindness and vision impairment; GBD 2019 Blindness and Vision Impairment Collaborators, Lancet Global Health 2021 (43.3 million people blind in 2020).
Eight NASA datasets turned into sound, era by era. Pick a year, load a tape, listen to the planet, or click anywhere on the globe and hear that patch of Earth.
Every cassette holds one dataset for one era: GISTEMP surface temperature, SVS 5196 model carbon dioxide, AMSR2 sea ice, MODIS snow cover, IGBP land cover, IMERG precipitation, MODIS aerosol and net ecosystem CO2 exchange.
When a tape is inserted its source arrives once, as a CSV column, an animation frame or a cached satellite tile, and is resampled to a 512 × 512 frame matrix. That image is read as a spectrum: each column becomes a chord, each brightness value a harmonic. The result is rendered into an audio buffer in the browser, with no backend and no pre-recorded files.
Eight tapes exist in total, but a rack only carries the ones whose data reaches that era: 1990 has two, 2010 has six, 2025 has all eight. Drag the fan or use the arrows to reach one, click it once to centre it and once more to drop it into the deck.
Click a point on the sphere and the app loads the matching satellite tile (z/x/y), downsamples it to the same frame matrix and sonifies it.
v2, spectral. 16 partials per column, glide between frames, 4.8 kHz low-pass, mono. It reads like a spectrogram played left to right: an X-ray of the image.
v3, musical. The same matrix is quantised to an A-minor pentatonic ladder, capped at 6 partials, softened attacks and stereo spread. Same data, but it sounds like an instrument instead of a scan.
Switch with the motor button in the globe HUD; the tape deck always uses the musical engine.
Eight topics sit on a draggable strip: temperature, aerosols, sea ice, forests, carbon, snow, rain, flux. Each card opens a dossier with the numbers behind the sound: what was measured, by which mission, and what the trend means.
Dossiers are built from the same source files, so the numbers on the page always agree with what you just heard.
The gallery is a cylinder of eras, 1990 → 2025. Clicking a capsule loads that era's tapes into the deck, so you can hear how the same measurement changes across three and a half decades of satellite records.
Every cassette and every globe tile comes from a public NASA source. Links go straight to the dataset or product page.
Sonification: additive synthesis written for this project, v2 spectral engine and v3 musical engine, both running in the browser with Web Audio.
Short answers about the data, the sound and what runs where.