Yohta Kitagawa

NoiseLoom

Yohta Kitagawa · 2026 · Solo project

An accessible interface for improvising noise and finding the sound together.

An accessible interface for improvising noise and finding the sound together.

NoiseLoom is a networked instrument for collective noise improvisation. Participants scan a QR code and draw on a shared spectral canvas from their phones. Their marks overlap on a projection and become one continuously changing sound.

Everyone edits the same sonic material rather than playing a separate instrument. Participants listen to the collective result, see each other's actions, and respond by adding, erasing, blurring, or transforming the canvas.

Technical Details

  • Phone browsers send touch and color data over WebSocket to a Node.js server.
  • The server forwards OSC data to Max/MSP.
  • A shared spectrogram drives spectral synthesis: position sets time and frequency, brightness sets volume.
  • Image transformations (blur, erase, shift) reshape the sound.

Publications & Exhibitions

  • Poster submitted to ACM UIST 2026.
  • Exhibited at BLIP, Brown University (May 13, 2026).

Process

NoiseLoom began with a question: how can people improvise with complex sound together without specialized software? A browser-based interface answered it, letting anyone join from their own phone within seconds.

Turning the shared spectrogram into both an image and an instrument made the collective nature of the work explicit. Edits accumulate rather than replace one another, so the sound stays a construction that no single participant owns.

A preliminary study with eleven participants reported a strong sense of social connectedness, while newcomers felt less confident shaping the sound. That gap now guides the work toward clearer feedback, a wider range of sound materials, and further spatial-audio experiments.

Credits

Created by Yohta Kitagawa. Originally created in a class at Brown University. Thanks to Professor Butch Rovan. Videography by Ingrid Jarvi.