SoundSpotter is a tool for organizing and retrieving sound grains in real time, letting performers steer through audio-similarity space with a handful of interactive controls.
It is a highly-optimized C++ package that cross-compiles on Windows, OSX and Linux, with an API for computer-music environments such as Max/MSP, PureData and SuperCollider (the first two implemented through Thomas Grill's FLEXT package). SoundSpotter was written by Michael Casey starting in 2000, with the first open-source version released in 2004. Its search engine grew into audioDB, part of the OMRAS2 project.
Rather than synthesising new sound, SoundSpotter matches: given a live or recorded target, it continuously finds the closest-sounding grains from a source corpus and plays them back. The result is a kind of real-time musical mashup, and a practical front-end for content-based music information retrieval.
News & Notes
Our News & Notes section is written by a small editorial team of computer-music and MIR enthusiasts who document the software, the research around it, and its use in performance.
Esben Rask
Audio & DSP
Esben writes about the signal-processing side of SoundSpotter: feature extraction, matching, and getting a clean build going on OSX, Windows and Linux. He has spent years around computer-music toolkits and live electronics.
Lucia Mendel
Music Information Retrieval
Lucia covers the research behind audio matching, from acoustic lexemes and shingles to approximate nearest-neighbour search. She is drawn to the places where MIR meets listening and composition.
Teodor Vasiliev
Sound, Performance & Archives
Teodor documents SoundSpotter in performance: patching it into Max/MSP and PureData, steering through similarity space on stage, and keeping the project archive and TimbreScores gallery in order.
Dorian Vale
Guest contributor
Dorian Vale is a guest writer on sound design and audio in interactive media.
Questions or corrections? Get in touch.