WorkEdge and cloud
SetSlice
Desktop software that turns one long recording into separate, named songs, even when there's no gap between them.
- ~5.5 KB
- sent per song lookup: a fingerprint computed on your computer, never the audio. About 300 KB per hour of recording
- 0.2 s
- median gap between a cut and the song start its fingerprints point to; all 10 cuts in a 48-minute test within 1.2 s
- ~3 min
- to name every song in a 48-minute recording; the local signal analysis takes 9 seconds of that
- 25 hours
- of song identification in the one-time early-access price; splitting and MP3 export run on your computer
SetSlice started as an in-house tool. Long, continuous sessions of music recorded in Audacity needed to come back as separate, properly named songs, without an evening of dragging markers.
The usual approach, cutting at the silence between songs, fails when songs run straight into each other. SetSlice listens instead. It fingerprints a short clip about every seventy seconds and sends only that fingerprint to a recognition service, and each match reports where in the original song that clip sits. That gives the song’s start time directly, and separate clips of the same song agree to within milliseconds. Local signal processing then places each cut at the quietest point near that estimate, and a consistency check throws out matches that contradict better-supported ones, which matters when one track samples another. The recognition provider sits behind a small interface, so it can be swapped.
It runs on the desktop and drives Audacity through its scripting pipe. It reads the audio, adds a labelled track so every cut can be checked before anything is saved, then exports each song once as a tagged MP3, encoding several in parallel. Splitting happens locally; only short fingerprints leave the machine.