bróga is a generative composition engine that treats language as musical material. It queries the Datamuse lexical API in real time — pulling words by phonetic similarity, semantic association, and rhyme — and routes that material through a self-modifying grammar of eight compositional rules. The result is a piece that is never the same twice and never entirely its own author.
Each word that arrives carries a semantic tension value: a measure of how far the current vocabulary has drifted from its origin. That tension drives the synthesis directly — shaping the grain cloud density, pulling the vowel formant field across the spatial plane, bending the interval grammar toward or away from crystallinity. When tension peaks, the system can mutate its own rules. When it collapses, it crystallises into temporary order before dissolving again.
Sound is constructed entirely from Web Audio primitives: Karplus-Strong plucked strings, FM choir voices tuned to live formant ratios, a granular engine seeded by the phonetic body of each incoming word, drone oscillators tracking the root, and a spatial field placing every voice in a two-dimensional listening plane. Nothing is sampled. Everything is calculated at the moment it sounds.
The score renders in real time as a custom glyph notation — each phrase a mark, each rule a shape, each silence a gap. It can be exported as MusicXML for further notation work, or as a PDF graphic score.
This experience uses Web Audio synthesis and a live score renderer that require a desktop browser. Please visit on a laptop or desktop to continue.