Observation
Signal processing measures the sound continuously: frequencies, harmonics, spectral energy, entropy, sudden transients and persistent resonances.
Experimental · in development
An experimental system for continuous computational listening.
Listening Observatory connects to live audio streams, analyses their acoustic structure in real time, and translates what it hears into evolving visual forms and short textual observations. Rather than treating sound as something to be displayed moment by moment, it listens across timescales: from immediate frequency changes to recurring resonances, motifs, and longer patterns it remembers.
Signal processing measures the sound continuously: frequencies, harmonics, spectral energy, entropy, sudden transients and persistent resonances.
A memory compares what it hears with what it has heard before. Structures that recur become motifs with a persistent identity, and are recognised when they return.
Drawn live in the browser from these observations: a landscape of the spectrum, resonances rising from it, disturbances where transients strike, and a dial of what has been remembered.
A language model phrases the measured facts in two short sentences. It never decides what happened in the sound: every frequency and motif it names is checked against the measurements, or the text is discarded.
The DSP system observes · the memory recognises · the visualization embodies · the language model gives words to what was measured.
Listening Observatory is open source (github.com/vrontier/listen). The audio belongs to its sources and is credited on each stream's page. Questions, ideas or a stream worth listening to? Get in touch.