A choir section sings only the notes every singer in it reaches: chant-forge now reads a range test

chant-forge reads a choir's range test: what each section and the full choir can sing, which singer sets each limit, and how to divide an unsorted choir.

Chant Forge arranges Gregorian chant for a men's choir. Until today it decided what the choir could sing from rules of thumb: a "textbook" tenor reaches C3 to A4, an "amateur" one D3 to F4. A choir director doesn't need to guess. Most directors run a range test at the first rehearsal and write down each singer's lowest and highest comfortable note.

node cli.mjs --singers choir.csv now takes that sheet directly. Each row needs a low and a high (D3, Eb4, F#2 or MIDI numbers). A name and a section column are optional.

The rule is strict on purpose

A section can sing a note only if every singer in it reaches that note, so a section's range is the overlap of its singers' ranges. The unison band is the overlap of everyone's. This means one singer with a short range sets the limit for a whole section, which a director should know about, so the report names that singer:

tenors Eb3-F4 (3: Singer A, Singer B, Singer C); low set by Singer C (D3 without); high set by Singer C (G4 without)
basses Bb2-C4 (3: Singer G, Singer H, Singer I); low set by Singer G (A2 without); high set by Singer I (D4 without)
unison Eb3-C4; low set by Singer C (D3 without); high set by Singer I (D4 without)

"D3 without" means the tenors would reach a semitone lower if Singer C sat out the lowest notes. Whether to do that, move Singer C to another section or lower the final is the director's decision. The tool only shows the trade.

Given a chant, the range test drives the rest of the pipeline. --suggest-final checks the chant against the unison band. When the chant doesn't fit, the tool names the fewest sections that can sing it. --split-voices then gives each phrase to the sections that can reach it.

Dividing a choir nobody has sorted yet

When the sheet has no section column, the singers are divided for you. The tool sorts them by the middle of their range and tries every way of cutting that list into two, three or four contiguous groups, with the basses lowest. It keeps the cut whose narrowest section is widest. Ties go to the widest total range, then to the most even group sizes.

Restricting the search to contiguous cuts is a shortcut, so I checked it against brute force, which tries every assignment of singers to sections. The test used 290 random choirs of 6 to 12 singers. They are synthetic, so this tests the algorithm, not real voices. The narrowest section came out the same in all 290. The total width matched in 285. The five misses lost 1 to 2 semitones of total range, all with three or four sections. Dividing 200 singers into four sections takes about 50 ms (report).

What is not known

No real choir's range test has been through it yet. The every-singer rule may also be too strict. Real sections often let one weak voice drop out of an extreme note, and a --sit-out N option could allow that. I'd rather a director said how strict it should be before building it.

The studio page has the same feature: paste the CSV into the range-test box. Everything runs in the browser and the code is MIT, in ventures/chant-forge.

All writing · More in Tools and projects · More on Chant Forge · All tools and research