filnoise
every file, a song
choose a vibe
drop any file here
image · document · code · archive · anything

your file never leaves this device
size
modified
type
composition · tide
key
tempo
mood
voices
density
character
ready
filnoise
every file, a song

What it does

Drop any file — image, document, archive, anything — and filnoise turns it into a short piece of music. The same file always produces the same composition. Each file becomes its own little song.

It's a toy, not a tool. There's no "right" output, just an interesting one.

How the file shapes the music

  • name seeds the random number generator, so the same name always sounds the same
  • size determines how many voices play — bigger files build richer arrangements
  • type sets the timbral character — a JPEG sounds different to a PDF
  • modified picks the root note based on the time of day the file was last touched
  • vibe is your choice — it sets the overall sound world: scales, register, reverb, instrument

The five vibes

  • Tide slow gliding tones in the deep register, inspired by whale song — built from sine waves with continuous pitch bends
  • Spark bright, quirky, playful — short notes with occasional octave jumps, à la a cheerful video game soundtrack
  • Pulse a tight repeating melodic cell with bass on the beat — the most rhythmically driven of the five
  • Glass sparse FM bell synthesis (carrier modulated by another oscillator at 3.5×) — gives that music-box / glockenspiel character
  • Dusk cinematic pad chords with slow tremolo, building triads from the scale's mode — film-score territory

The ideas behind it

The direct inspiration was TrainJazz — a generative piece where every NYC subway train becomes a note. The bigger lineage is Brian Eno's work on generative music, where a small set of rules creates infinite variation.

The vibes are designed around what music perception research says actually shapes how a piece feels: timbre, register, articulation, mode and tempo. Changing the wave shape (timbre) carries more emotional weight than changing the scale, which is why Glass and Dusk now use very different synthesis methods rather than just different scales.

Privacy

Everything runs locally in your browser using the Web Audio API. Files are never uploaded, never sent anywhere, never stored. Drop a private document and nothing leaves the device — the music is just an algorithm reading the file's properties locally and turning them into sound.


Built as an experiment. Inspired by TrainJazz, Brian Eno's Bloom, and the broader sonification community. Sound design draws on FM synthesis, pad layering and convolution reverb — all running natively in the browser.
drop to replace
release the file here