Aurelia Overture and Velvet Circuit cannot render: their Chorus racks predate the effects rebuild

[bdbarnett]

Aurelia Overture and Velvet Circuit both build a Chorus the way audioeffects used to take one, and neither piece renders any more:

$ composer/preview.py --piece AureliaOverture out.wav
TypeError: Chorus._build() got an unexpected keyword argument 'depth_ms'

Aurelia dies three tracks in, on String Hall; Velvet Circuit dies on Neon Suitcase. Both are the first rack in their piece that reaches for a chorus, so the failure is the API and not the music.

audioeffects.Chorus is now:

_build(self, rate=0.8, depth=0.35, mix=0.815, delay_ms=8.8, tone_hz=3000.0, patch=None)

The racks pass depth_ms=, voices= and width=. depth is the survivor of depth_ms; there is no per-voice count and no width any more. It changed in audiocomponents cafcc57, "Phase 3: five modulation effects rebuilt, one dropped, two carried forward" — the compositions were not moved with it.

One more waiting behind it, found by reading rather than by running, because the Chorus stops the render first: Velvet Circuit's Ghost Projector builds Vibrato(source, rate=4.7, depth_semitones=0.06) and today's Vibrato._build takes depth, not depth_semitones. Everything else in both pieces checks out against the current signatures — TapeDelay still absorbs feedback/mix/wow/tone_hz through **options, and Saturation, Overdrive, Reverb, SlapbackDelay and Phaser all still take what the racks give them.

Why it is worth fixing rather than leaving

These are not just previews. pieces.py::load_piece is what reaper.py, verify_song.py and reaper.sh all go through, so neither piece can be played, generated or bounced either.

They are also two of the seven pieces in audioif's soundtrack render gate (tests/parity/capture_render_reference.py), which I have just repointed at this renderer for audioif#88. The gate now runs and grades six of eight pieces; these two come back failed, so a DSP change in audioif that moved Velvet Circuit's eleven hardware emulations would go unnoticed.

Note that a rack is a patch, embedded byte-for-byte in a generated .RPP — so changing these numbers changes how the piece sounds, and picking the new values is a musical decision, not a mechanical translation of the old ones.

Related: audiocomponents#83, a second seam where the soundtrack and today's audiocomponents disagree.

[bdbarnett]

Arthur, 2026-09-18: not a mechanical fix, as the issue says — the new Chorus and Vibrato values are a musical choice, so this is the effects program's Phase 6 soundtrack migration (audiocomponents#34), where every piece moves onto the rebuilt surfaces and is re-captured. Brad has ruled no relistens until the effects work lands, so it waits there; until then the render gate grades six of eight pieces and says so. audiocomponents#83 (the other seam) is fixed on effects/p4-integration.