A ninth-century second voice for the repeats: organum in Chant Forge

Chant Forge's --repeat sang the same unison line every pass. Now a Musica enchiriadis organal voice sings under every pass after the first.

Chant Forge turns Gregorian chant into tracks for the gym: an all-male choir over deep house, EDM, hard rock or metal. A one-minute antiphon becomes a three-minute track with --repeat 3. Until today that meant the choir sang the same unison line three times. The backing changed from verse to drop, and the voices did not.

The fix comes from inside the tradition. Around 900, the treatise Musica enchiriadis wrote down the first way of adding a second voice to chant. Its rules are short and mechanical, so they fit in about sixty lines of JavaScript.

The rules

The chant is the vox principalis. The added vox organalis sings it a fourth below, with three limits:

  1. A boundary. The added voice does not go below a fixed note under the final: C under a D final. Where a fourth below would go lower, it holds that note. As the chant comes down to its final, the gap shrinks from a fourth to a third, a second and then a unison. This narrowing is the style's best-known sound.
  2. No tritone. Where the fourth below is not in the chant's scale (B natural under an E, when the chant only uses B flat), the voice takes the third below instead.
  3. Occursus. Every phrase ends with both voices on the same note.

Here is the second pass of Dies irae (D3 final, boundary C3), chant over added voice:

Di-  es   i-   rae   di-  es   il-  la
F3   E3   F3   D3    E3   C3   D3   D3
C3   C3   C3   C3    C3   C3   C3   D3

The added voice sits on C while the melody moves above it, then joins it on the last note. That matches the treatise's own D-mode examples.

What it does in the tool

--organum leaves the first pass of a --repeat as plain unison chant and adds the second voice from pass two on, so the track opens with chant and the harmony arrives later. Without --repeat the second voice runs under the whole piece. It goes on its own MIDI track, "Organal voice", with the Latin lyrics. So the preview renderer plays it, the sung preview sings it, and the rehearsal-track export gives it a part track and a minus track like any other section. The studio page has a "Second voice" switch.

Range is checked. If the added voice would drop below E2, the command stops and names the lowest final that fits. With a key lift on the last drop, the boundary moves up with the lifted line.

Measured on the corpus

Over all 68 GABC files in the corpus, each at the key the tool would pick for a men's choir, the added voice sings 14,079 notes:

interval under the chantshare
fourth51.9%
major third15.6%
minor third4.4%
second13.5%
unison14.6%

It never sings a semitone or a tritone under the chant, and it stays inside E2-E4 in all 68 files. One rule changed during the run: the first version set the boundary at the scale step just below the final. For an F final that step is E, a semitone down, which put a minor second under every F (2.6% of all notes). The boundary now has to be at least a whole tone below the final, and the semitones are gone.

Without --organum the output is unchanged: four sample commands produce byte-identical MIDI and cue sheets. There are 10 new tests, and the browser check passes all 127 of its checks.

What nobody knows yet

Nobody has listened to it. A fourth below a melody might sound medieval over a sub-bass, or it might sound muddy. Singers who learned the line in unison might or might not hold the organal part. The tool can't answer either question; a listener and a choir can.

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