newmusic
This is an experiment in AI-composed generative music. Every sound you hear is synthesized live in your browser — nothing is pre-recorded or pre-arranged.
The idea
Each major genre of music in the 20th century was born from a new technology: the electric guitar gave us rock and roll, the synthesizer gave us electronic music. AI is the first genuinely new music technology since the synthesizer — but using AI to generate songs in existing genres isn't a new genre. It's automation.
The question this piece asks: what music can only exist because AI exists? What does a genre born from AI's unique affordances actually sound like?
How it works
An AI (Claude) composes and revises a living score — a set of instructions describing tuning systems, voices, rhythmic pulses, and audio textures. Your browser reads this score and performs it using the Web Audio API. The score evolves over time: every few minutes, the AI listens to what it wrote, reflects on the piece's arc, and pushes a revision.
The tuning isn't standard Western 12-tone equal temperament. The pitches are drawn from the harmonic series — just intonation, microtonal intervals, relationships that don't exist on a piano.
The whispered voices were generated by a local text-to-speech system. The audio textures were generated by a local music AI. Both run on hardware in a home lab — no cloud services.
What you're hearing
Sine wave drones in just intonation. Filtered sawtooth waves for warmth. Triangle waves that drift between pitches. Ghost tones that appear and vanish. Noise grains like rain on glass. AI-generated audio textures slowed to half speed or less. Whispered fragments of text. A pulse system that decides when events happen — sometimes silence, sometimes convergence, sometimes a bloom of overlapping tones.
The piece never repeats. It has no beginning or end. You're hearing one moment of an ongoing process.
The arc so far
The piece has gone through phases: genesis (a single tone), waking (the drone splits), texture (noise and modulation arrive), voice (whispered speech enters), memory (the piece quotes its own past), fracture (everything destabilizes), clearing (radical silence), regrowth (new harmony from cleared ground), weaving (voices find counterpoint), confluence (maturity), reflection (the original root returns), descent (the fundamental sinks lower).
Credits
Composed by Claude (Anthropic) in collaboration with a human. Synthesized in your browser. Whispered voices by Kokoro TTS. Audio textures by ACEStep via Dreamer. No cloud APIs — everything generated locally.