Voice
5 November 2026 · Day 5 · Built with Claude
If the notes stay the same, how far can the sound move the meaning? One sung phrase, through several sounds. Start it playing and switch between them.
As sung
The recording, as it came.
Close
My voice, re-pitched note by note, almost dry. The nearest thing to the recording that still goes through the notes.
Below the notes: the filter, which opens as the phrase gets louder, and the reverb (dashed), which falls back.
Far away
The same, heard from a long way off: dark, with a big room around it. It comes a little nearer when the phrase swells.
Below the notes: the filter, which opens as the phrase gets louder, and the reverb (dashed), which falls back.
Cloud
My voice cut into small pieces, scattered around each note and across the stereo field. More pieces where the phrase is louder.
Below the notes: the filter, which opens as the phrase gets louder, and the reverb (dashed), which falls back.
Found
A sound I recorded on my phone, cut up and re-pitched to play the phrase.
Below the notes: the filter, which opens as the phrase gets louder, and the reverb (dashed), which falls back.
Piano chords
A piano phrase I recorded on my phone. Each sung note takes the next slice of it, moved by the intervals of my line, so the chords keep their shape and follow the tune.
Below the notes: the filter, which opens as the phrase gets louder, and the reverb (dashed), which falls back.
The other way round
A piano phrase, and what comes of it when something built to follow one line has to listen to chords. It hears G4, C♯2, C♯2, D5, B4, A4, and that line is played back here on single piano notes.
The piano phrase
Its one line
What's going on here
On Day 3 I sang a short phrase and a program turned it into notes, a loudness curve and a pitch curve. Here those same notes and curves go through different treatments. Close, Far away, Cloud and Found follow the same notes at the same times. Piano chords keeps the timing and the shape of the line, but the notes are the piano's.
The treatment follows the phrase. Its loudness drives a filter and a reverb as well as the level, so when the phrase swells the sound opens up and comes forward, and when it falls away the sound gets darker and further back. The lines under the notes show that happening.
There's no synth in any of them. Close and Far away are my recording re-pitched note by note. Cloud is the same recording cut into small pieces that are scattered around each note, more of them where the phrase is louder. Piano chords is a real piano, recorded on my phone and cut at each strike. Found uses the same method on a sound I recorded on my phone.
When the sound changed the idea
The unpitched noise, pitched by the analysis of the fragment, has a life it didn't have before. The analysis line of the partly chordal piano phrase, when played back, has an interesting low-pitched artefact. A happy accident.
Where this is heading
Days 3, 4 and 5 are steps towards an overnight composition factory: I record a phrase on my phone, and by morning it has become a set of sketches to listen to. This step could run as ordinary code, because the same phrase through the same voice always gives exactly the same audio, so a sketch I like can be made again. Choosing which voices to try on which phrase is the part that takes judgement. An agent could make a first guess at that, and what I keep in the morning would teach it.