Chris ColatosChris Colatos

In Suspended Harmony

A 6:47 composition for 20 DSS voices revealing the overtone series of a single B♭ fundamental. Partials 1 through 16 emerge across nine harmonic regions, building from a bare octave to full spectrum before dissolving back to the fundamental, altered by a ghost of the 15th partial.

Lemniscate

A 6:47 realization of DSS with no key, no scale, and no harmonic plan. Twenty voices are defined only by frequency barriers in raw Hz. The random walk determines every pitch. Consonance is a statistical event.

Gendy c²

A 6:47 sparse realization faithful to Xenakis's GENDY3. Only one to five voices play at any moment, exposing each waveform mutation as a naked, audible event rather than burying it in texture.

Chiaroscuro

An 8:08 composition modulating through six tonal centers (B♭, F, E♭, Gm, Dm, A♭) across twelve regions. Four voice classes rotate leadership, eleven asymmetric breath-swells shape the dynamics, and Cauchy bursts punctuate harmonic shifts as structural events.

Ergodic

A 9:37 composition built on durational structure rather than harmonic structure. No tonal center. Twenty voices span 18 Hz to 14 kHz, untethered from harmonic function; frequency spread itself becomes compositional material, narrowing to near-pure tones in silence and exploding beyond defined ranges during twin eruptions. Twelve sections follow Fibonacci-adjacent proportions across 577 seconds (prime). Stochastic parameters reach extremes no other piece in the collection touches: full Cauchy distribution, maximum walk rates, density from a single sub-bass voice to all twenty at once. The form is symmetrical: void to void, with two violent climaxes separated by a shattering collapse.

All five compositions share a single engine: a recreation of Xenakis's Dynamic Stochastic Synthesis in a Web Audio API AudioWorklet. Breakpoints evolve at sample rate via Gaussian and Cauchy random walk with reflecting barriers. Stochastic parameters are themselves subject to second-order random walk. Signal chain: DSS voices, tanh saturation, 12-second convolution reverb, shimmer delay. No samples. No oscillators. Every sound is generated from arithmetic on breakpoint arrays.

Browser-based generative composition engine that writes voiced, phrased, and cadenced music across jazz, classical, electronic, and song idioms, builds the rhythm under it, then lets you reshape it. Five linked views: a piano-roll Composer Map with chord cards and a live harmonic-tension trace, a Harmonic Orbit of chord substitutions ranked by shared tones, a Voice Loom for voice-leading edits, a Field that lays the generative layer across the whole passage, and a Drum Room whose break engine writes the groove across more than fifty styles, steered by chaos, ornament, and polymeter. Temperament sliders for risk, color, motion, and spread, six one-tap harmonic transformations, and motif-based melody development. Two traditions of algorithmic composition run side by side. Learn is symbolic AI in the lineage of David Cope's Experiments in Musical Intelligence: it ingests your own MIDI, infers key by Krumhansl-Schmuckler profile correlation, separates voices by onset clustering, distills recurring motifs, and builds a first-order Markov model of chord-root motion, then composes fresh music in that style by rule rather than imitation. Process is its formal counterpart, six layered generative engines: Euclidean rhythm, cellular automata, L-systems, Markov chains, Xenakis-style stochastic distributions, and parametric curves, each with its own live visualizer, every note locked to the sounding chord. Multi-track Standard MIDI File export, OfflineAudioContext WAV rendering, and sample-accurate scheduling on the Web Audio clock. Route live Web MIDI, each part on its own channel, to your synths or a DAW to hear it at full resolution.

Browser-based generative drum instrument that fuses sample-based breakbeat manipulation with algorithmic pattern synthesis. STFT-based onset detection and adaptive classification split any uploaded break into kicks, snares, ghost notes, and hats. A chaos engine performs progressive structural transformations through per-voice step swaps, half-bar mirroring, bar rotation, cross-voice substitution, and stutter injection, exposing the instrument's compositional logic as a continuously tunable parameter space. Twelve time signatures including odd and compound meters. Eleven-voice mixer with solo/mute, pan, faders, and real-time VU metering. Sidechain ducking, comb filter, bitcrush, 4x oversampled tanh drive. Click-and-drag velocity editing per cell. Standard MIDI File export, OfflineAudioContext WAV rendering, live Web MIDI output. Sample-accurate scheduling locked to the Web Audio clock.

Nine-channel tracker-style step sequencer and synthesizer running entirely in the browser. No samples. Every kick, snare, bass, and pad is synthesized live from oscillators, noise, and waveshaping via the Web Audio API. Five drum and four melodic channels, each with multiple synthesis variants, variable time signatures including odd meters, grid resolutions to 1/256, swing, delay, and convolution reverb. Per-note probability and Elektron-style trig conditions. Pattern variants (A/B/C/D) assignable per bar. A preset library of 135 rhythms across 18 genres, from the jungle breaks to breakcore, techno, IDM, and artist references (Aphex Twin, Venetian Snares, Autechre, Squarepusher). Copy/paste bars, per-channel pattern shift, one-level undo, and autosave to browser storage. Sample-accurate scheduling locked to the Web Audio clock, Web Worker-driven background playback, Standard MIDI File export, and live Web MIDI output to external hardware and software. Two UI skins, including a retro OctaMED tribute.

Sound made visible. A single-file browser instrument in three journeys. Foundations: thirteen stations for anyone starting out, from frequency and amplitude, phase and timbre through a five-band parametric EQ whose curve is read back from the running filters, the ear's own sensitivity, masking and precedence, a frequency-recognition trainer that adapts to the listener, a sample-by-sample converter with audible aliasing and dither, gain staging with a computed noise floor, a room whose modes and boundary notch follow its dimensions, a compressor, polar patterns with the proximity effect derived from the pressure gradient, reverb and width on a goniometer, and loudness metering with exact ITU-R BS.1770 K-weighting; your own audio or microphone can be the source. Ratio: eight stations on interval, resonance and the ear, from Lissajous figures drawn from the exact phases of the sounding tones to a Matter station that solves the free square and circular Kirchhoff plates from first principles, matching the published tables to four digits, with powder settling on the nodal lines and Faraday waves in a dish of water. Synthesis: nine playable stations from the subtractive voice through FM with Bessel sideband prediction, wavetable, additive, granular, and a Karplus-Strong string in an AudioWorklet, played from the screen, the keyboard, or Web MIDI. Plain-language and physics notes on every station, audio that keeps time in the background, no samples, no dependencies.

Music theory for people who make music. A single-file browser app in thirteen stops that follow a producer's path: Pitch, Intervals, Scales, Modes, Pentatonic, Triads, Sevenths, Progressions, Beyond (borrowed chords, secondary dominants, passing diminished), Melody, Rhythm, Form, and Ear. Every stop pairs a playable keyboard and piano roll with demonstrations you can stop and restart, loops that keep playing while you change the tonic, scale, voicing, chords, or layers, and forty-three adaptive drills across 121 levels that step up and down with your streak. Whatever is sounding lights the keys, from a demonstration, a loop, a drill, or your own hands, on the screen, the computer keyboard, or Web MIDI with per-device sustain, octave stacking, and scale lock. Keys, pad, bass, and drums are synthesized in the browser (no samples), with click-free envelopes, a transport that swaps patterns at the next beat, and tempo changes quantized to the beat. Melody is a sixteen-step roll with motif operations (repeat, sequence, invert, shift) and live non-chord-tone analysis. Rhythm is a step sequencer with hi-hat subdivisions down to sixteenth triplets, swing on either grid, humanize, bass line, polyrhythms, and eleven genre beats with ghost notes and open hats. Form is a section timeline with layers, fills, risers, and silence before a drop, played as a preview or at full length. A Reference holds the circle of fifths, key signatures, intervals, chord formulas, modes, progressions, drum charts, and a glossary. Screen-reader key mirror, keyboard-operable grids and circle of fifths, reduced-motion support.

A playable recreation of the 1978 arcade original, with every sound left for you to make. Thirty-one game events, from the four-note alien march and the saucer's loop to pause, menus, and a shot that reaches the top of the screen, each with an empty slot: name a file after the event, drop it in, and it plays in the game in real time. The ten sounds the original board actually had are marked; the four events it left silent are marked too. Rehearsal controls summon a saucer, clear a wave, or take a life so any moment can be heard on demand, and a sound check auditions every event in order. A one-bar step sequencer voiced with square waves and noise supplies the music: pattern A for winning, pattern B for losing, with the game's tension deciding which plays, how much percussion is heard, and how fast, switching only on the bar line. Per-event volume in dB, a peak meter with clip hold, and sound packs that save the whole set as one file.