From fa9e88fb41c89e8fedfcab360723edcbe4d54811 Mon Sep 17 00:00:00 2001 From: Victor Giers Date: Tue, 14 Jul 2026 00:22:02 +0200 Subject: [PATCH] Add documentation for EP-133 project: Neon Laundromat --- .../IMAGINARY_PROJECT_NEON_LAUNDROMAT.md | 376 ++++++++++++++++++ docs/research/ep133-musician-audit/README.md | 2 +- 2 files changed, 377 insertions(+), 1 deletion(-) create mode 100644 docs/research/ep133-musician-audit/IMAGINARY_PROJECT_NEON_LAUNDROMAT.md diff --git a/docs/research/ep133-musician-audit/IMAGINARY_PROJECT_NEON_LAUNDROMAT.md b/docs/research/ep133-musician-audit/IMAGINARY_PROJECT_NEON_LAUNDROMAT.md new file mode 100644 index 0000000..2cd2127 --- /dev/null +++ b/docs/research/ep133-musician-audit/IMAGINARY_PROJECT_NEON_LAUNDROMAT.md @@ -0,0 +1,376 @@ +# Imaginary EP-133 project: Neon Laundromat + +`Neon Laundromat` is a dusty broken-beat/dub track made from drum-machine hits, Rhodes, bass, a spoken phrase and field recordings from a laundromat. It is meant to feel like a musician's project first and a coverage fixture second. + +The four-group split—drums, bass, harmony and vocals/textures—is conventional enough to remain playable on the hardware. The unusual details are things an adventurous KO II user could plausibly accumulate while finishing the track. + +## Identity + +| Setting | Value | Musical reason | +|---|---:|---| +| Firmware assumption | EP-133 OS 2.5.1 | Current behavior, including project key/scale and resampling improvements | +| Tempo | `112.50 BPM` | Broken-beat pace with space for swung percussion | +| Key | `D` | Tonal center for bass and Rhodes | +| Scale | `Dorian` | Minor color with a brighter sixth | +| Swing | `57%` at the 1/16 interval | Hats lean without making the kick unstable | +| Default meter | `4/4` | Main groove | +| Approximate song length | `216 quarter notes`, or `1:55.2` | Derived from the song-position list, including meter changes | + +Pad IDs such as `A1` and `A12` below are human-facing logical pad numbers for this design document. They are not claims about `pads/a/pNN`, SysEx pad IDs or the printed keypad glyph order. A real hardware fixture must record the physical-pad-to-TAR mapping explicitly. + +## Saved mix and output concept + +### Current selected send FX + +The final saved project has one selected **Delay**: + +- X / length: about `35%`; +- Y / feedback: about `52%`; +- Group A static send: `10%`; +- Group B static send: `4%`; +- Group C static send: `28%`; +- Group D static send: `38%`. + +Several group patterns contain recorded `FX` fader motion for delay throws. Earlier Filter and punch-in performances were resampled into D11, so they are not competing “second master effects” in the final saved project. + +### Output dynamics + +- Output compressor DRIVE: about `68%`; +- Output compressor SPEED: about `32%`. + +### Note-triggered sidechain + +- Source pads: audible kick A1 and silent trigger A12; +- Destination groups: B, C and D; +- LENGTH: about `42%`; +- SHAPE: about `65%`. + +A12 must duck the destinations even though it produces no source audio. + +### Runtime-only group project volume + +During the final live performance the musician sets the separate held-group project volumes to A `100%`, B `90%`, C `82%`, D `88%`. + +This is intentionally *not* part of the archived sonic oracle. The official 1.1.2 release notes say group project volume resets when a project loads. A backup-based exporter cannot infer the temporary values at which the musician most recently listened or performed. + +## Sound and sample provenance + +| Slots | Origin | Important property | +|---|---|---| +| 041–370 | imported and factory drum/percussion samples | mono one-shots plus a lower-rate shaker and a two-bar break | +| 412–430 | imported bass recordings | long tonal notes, a loop and one-bar sub drop | +| 501–570 | Rhodes, guitar, tape pad and organ recordings | mono/stereo, KEY chords, BAR/BPM stretch and a live-chopped phrase | +| 601–608 | spoken vocal and laundromat field recordings | one source phrase reused as several trims | +| 620 | internal stereo resample | punch-in/filter performance and live Delay X/Y movement baked into audio | +| 621 | resampled external polysynth | external MIDI result made self-contained as audio | + +Chopping is non-destructive in this project: several pads point to one source slot with different trim regions. Slot 620 and 621 are different: they are newly rendered samples whose creation history is no longer editable. + +## Group A — drums and sidechain triggers + +| Pad | Sample/slot | Musical use and pad-specific state | +|---|---|---| +| A1 | 041 `kick_tape.wav` | Main kick; ONE; AMP `100/200`; pitch `0`; centered; short trim; audible sidechain source. | +| A2 | **same slot 041** | Accent/click from the same kick; ONE; AMP `150`; pitch `+12`; only the first ~95 ms. This is the clearest same-WAV/different-AMP case. | +| A3 | 118 `snare_room.wav` | Main snare; ONE; AMP `108`; full body and room tail. | +| A4 | **same slot 118** | Reverse snare suction; REV; AMP `72`; pitch `-3`; pan L15; tail-only trim. | +| A5 | 205 `hat_closed.wav` | Closed hat; ONE; AMP `74`; pan L12; shared mute/choke group. | +| A6 | 213 `hat_open.wav` | Open hat; ONE; AMP `88`; pan R12; long release; same mute/choke group as A5. | +| A7 | 145 `clap_short.wav` | Clap layer; ONE; AMP `82`; deliberately late/free-timed. | +| A8 | 309 `shaker_soft_22050.wav` | 22.05 kHz mono shaker; ONE; AMP `66`; pan R28; recorded with pressure-sensitive note repeat. | +| A9 | 335 `conga_low.wav` | Low percussion; ONE; AMP `85`; pitch `-5`. | +| A10 | 370 `garage_break_2bar.wav` | Two-bar mono break; BPM mode from source `89.6 BPM`; AMP `48`; retriggered before the tail finishes in one fill. | +| A11 | 236 `crash_dirty.wav` | Crash; ONE; AMP `70`; long tail. | +| A12 | sound `000` | Silent sequenced pad used only as an additional sidechain note source. | + +Audible requirements: + +- A6 must be cut when A5 arrives. +- A2 must remain independent even though it references A1's WAV. +- A12 must duck B/C/D while remaining silent. +- A10 must retrigger according to EP ONE behavior rather than looping a Simpler region. + +## Group B — bass + +| Pad | Sample/slot | Musical use and pad-specific state | +|---|---|---| +| B1 | 412 `moog_D2_long.wav` | Principal bass; LEG; root D2; AMP `100`; very short attack; release about 180; played chromatically with overlapping notes. | +| B2 | **same slot 412** | Pluck variant; KEY; AMP `64`; pitch `+12`; first ~220 ms only; short release. | +| B3 | 414 `sub_D2.wav` | Clean sub; KEY; root D2; AMP `92`; long release. | +| B4 | 420 `reese_loop_90.wav` | Reese phrase; BPM stretch from `90 BPM`; AMP `78`. | +| B5 | 421 `bass_fret.wav` | Fret/noise accent; ONE; AMP `55`; pan R18. | +| B6 | 422 `bass_fall.wav` | Reversed bass fall; REV; AMP `80`. | +| B7 | 309 `shaker_soft_22050.wav` | Reuses A8's global slot, but trims one low transient, pitches it `-18` and treats it as bass percussion. | +| B8 | 430 `sub_drop.wav` | One-bar drop; BAR `1`; AMP `75`. | +| B9 | 412 `moog_D2_long.wav` | Soft-attack double of B1; KEY; AMP `128`; slower attack and shorter release. | + +B1 contains the LEG test: a D2 note remains held when an A2 note begins, so the pitch changes while playback continues from the existing sample position. B2/B3 contain overlapping KEY notes. + +## Group C — harmony and chops + +| Pad | Sample/slot | Musical use and pad-specific state | +|---|---|---| +| C1 | 501 `rhodes_Dm9.wav` | KEY; root D4; AMP `100`; simultaneous multi-note chords from one pad source. | +| C2 | **same slot 501** | Soft double; ONE; AMP `61`; fractional pitch `-0.09 semitone`; pan L24; trim begins slightly later. | +| C3 | 502 `rhodes_G13.wav` | ONE; AMP `92`; pan R8. | +| C4 | 503 `rhodes_Am7.wav` | ONE; AMP `88`; pan L8. | +| C5 | 504 `guitar_harmonic.wav` | REV; AMP `73`; pitch `+7`; pan R30. | +| C6 | 550 `tape_pad_stereo_32000.wav` | Four-bar 32 kHz stereo ambience; BAR `4`; slow attack and long release. | +| C7 | 551 `toy_arp_85bpm.wav` | BPM stretch from `85 BPM`; AMP `64`. | +| C8 | 560 `organ_stab_D.wav` | KEY; root D4; AMP `84`; short release. | +| C9 | 570 `rhodes_phrase.wav` | First live-chopped slice; trim 0–0.48 s; pan L30. | +| C10 | **same slot 570** | Second slice; trim 0.48–1.02 s; AMP `76`; centered. | +| C11 | **same slot 570** | Third slice; trim 1.02–1.63 s; AMP `69`; pitch `+5`; pan R30. | +| C12 | sound `000`, MIDI channel 3 | External-synth sketch pad, root D3. Pattern C90 drives a polysynth but is not used in the song list. | + +C1 contains a single-pad polyphony test: notes D, F, A and C overlap. C2 exercises the known negative fractional-pitch edge case. C9–C11 require separate playable instances of one WAV. + +## Group D — vocals, field recordings and printed performances + +| Pad | Sample/slot | Musical use and pad-specific state | +|---|---|---| +| D1 | 601 `stay_for_the_wash.wav` | Full spoken phrase; ONE; AMP `98`; centered. | +| D2 | **same slot 601** | “stay” chop; ONE; AMP `138`; pitch `-5`; pan L28. | +| D3 | **same slot 601** | “wash” chop; ONE; AMP `72`; pitch `+7`; pan R28. | +| D4 | 602 `breath.wav` | REV; AMP `58`; slow attack. | +| D5 | 603 `laundromat_crowd.wav` | Two-bar stereo ambience; BAR `2`; AMP `44`. | +| D6 | 604 `washer_cycle_83bpm.wav` | Rhythmic field recording; BPM stretch from `83 BPM`; AMP `52`. | +| D7 | 605 `riser.wav` | ONE; AMP `75`; long trim. | +| D8 | 606 `metal_door_impact.wav` | ONE; AMP `110`; pitch `-7`. | +| D9 | 607 `water_drip.wav` | ONE; AMP `68`; pan L35. | +| D10 | 608 `phone_beep.wav` | KEY; root A4; AMP `54`; pitch `-0.09`; pan R35. | +| D11 | 620 `printed_punchin_hook.wav` | Stereo two-bar resample of the hook while beat-repeat/filter punch-ins and live Delay X/Y changes were performed; BAR `2`. | +| D12 | 621 `printed_external_chord.wav` | Resampled external polysynth chord originally played from C12/C90; KEY; root D3. | + +D11 and D12 are provenance tests: Live should reproduce their rendered audio, but should not pretend to recover the gestures or external synthesizer that created them. + +## Pattern performance + +The project contains quantized notes, free timing, tick nudges, varied velocity and meaningful durations: + +- A03 has broken kicks both on and off the quarter-note grid, snares at beats 2 and 4 with different velocities, 1/16 hats with alternating velocities and several `-3/+5` tick nudges. +- A08's pressure-recorded note repeat becomes a shaker velocity ramp. The intended artifact is the resulting notes, not the player's finger-pressure history. +- A04 begins A6 before A5 arrives; the open hat must be choked by the later closed hat. +- A04 places A2 immediately before the final snare, followed by a short group FX-send throw. +- B03 overlaps B1 LEG notes: D2 remains held as A2 begins. Durations and velocities vary. +- C02/C04 use C1 for overlapping chord tones and layer independent C2 underneath. +- C05 contains a D-Dorian arpeggio originally performed with note-repeat/arpeggio; the resulting note sequence is what matters. +- D03 alternates D2 and D3 with different velocities and throws only the final word into Delay. +- D08 triggers D11 as a rendered two-bar fill while the other groups become sparse. + +Every playback mode has a meaningful test: + +- ONE: complete sample playback plus retrigger before the preceding hit ends; +- KEY: overlapping notes/polyphony from one pad source; +- LEG: held-playhead pitch changes. + +## Scenes and independent pattern lengths + +Human labels below are documentation annotations; numeric scene IDs are the project structure. + +| Scene | Label | Meter | A | B | C | D | Duration | +|---|---|---:|---|---|---|---|---:| +| 01 | Spin-up | 4/4 | A01, 4 bars | B01, 2 | C01, 4 | D01, 4 | 4 bars | +| 02 | Door shuts | 4/4 | A02, 2 | B02, 4 | C02, 4 | D02, 1 | 4 bars | +| 03 | Verse | 4/4 | A03, 2 | B03, 4 | C03, 4 | D03, 1 | 4 bars | +| 04 | Hook | 4/4 | A04, 4 | B04, 4 | C04, 4 | D04, 4 | 4 bars | +| 05 | Three-beat rinse | 3/4 | A05, 2 | B05, 1 | C05, 2 | D05, 2 | 2 bars | +| 06 | Dirtier verse | 4/4 | A06, 4 | B03, 4 | C03, 4 | D06, 2 | 4 bars | +| 07 | Missing-beat cut | 2/4 | A07, 1 | B07, 1 | C07, 1 | D07, 1 | 1 bar | +| 08 | Final spin | 4/4 | A08, 4 | B08, 8 | C08, 4 | D08, 2 | 8 bars | +| 09 | Power down | 4/4 | A09, 4 | B09, 4 | C09, 4 | D09, 4 | 4 bars | +| 10 | Dub lab | 4/4 | A10, 4 | B10, 4 | C10, 4 | D10, 4 | 4 bars; not in song list | + +Shorter group patterns repeat until the longest pattern in the scene ends. D03 therefore loops four times in Scene 03, while C08 loops twice in Scene 08. + +## Song positions + +```text +01 → Scene 01 +02 → Scene 02 +03 → Scene 03 +04 → Scene 03 +05 → Scene 04 +06 → Scene 05 +07 → Scene 06 +08 → Scene 03 +09 → Scene 07 +10 → Scene 04 +11 → Scene 08 +12 → Scene 08 +13 → Scene 09 +``` + +The song deliberately reuses scenes. Session View should expose reusable Scenes 01–10 once each. Arrangement View should follow all thirteen positions, including repeated Scenes 03, 04 and 08. + +The duration calculation is: + +```text +4/4 sections: 52 bars × 4 quarter notes = 208 quarter notes +Scene 05: 2 bars × 3 quarter notes = 6 quarter notes +Scene 07: 1 bar × 2 quarter notes = 2 quarter notes +total 216 quarter notes +``` + +## Fader automation in the main song + +- A01: HPF opens the intro from thin to full. +- A04: FX-send throw on the final snare. +- A06: VEL rises into the fill while preserving the differences among original note velocities. +- A08: LVL build plus final FX throw. +- A09: HPF and LVL close down the outro. +- B03: LPF opens gradually and settles. +- B05: PTC bends down into the 3/4 breakdown. +- B08: subtle TUNE drift over eight bars. +- C01: ATK falls from a soft swell to a clearer entrance. +- C04: FX-send accents chord endings. +- C05: REL grows through the breakdown. +- C08: HPF and FX move over the final hook. +- D03: short FX throws on vocal endings. +- D04: PTC tape-down gesture. +- D08: PAN moves the printed fill across the stereo field. + +These lanes belong to group patterns and repeat whenever those patterns repeat. + +## Scene 10: the retained Dub Lab experiment + +Scene 10 is a plausible musician's scratch/jam scene left outside the song list. It exercises all fader assignments without making the main arrangement absurd: + +| Group | Lanes | +|---|---| +| A10 | PAN, TUNE, VEL | +| B10 | PTC, TIM, MOD | +| C10 | ATK, REL, FX | +| D10 | LVL, LPF, HPF | + +Musician-facing envelope anchors over four bars: + +```text +A PAN: 0.25 → 0.75 → 0.25 → 0.50 +A TUNE: 0.50 → 0.58 → 0.42 → 0.50 +A VEL: 1.00 → 0.60 → 1.00 + +B PTC: 0.50 → 0.75 → 0.50 +B TIM: 0.50 → 0.80 → 0.50 +B MOD: 0.00 → 1.00 → 0.00 + +C ATK: 0.05 → 0.75 +C REL: 0.80 → 0.10 +C FX: 0.10 → 0.75 → 0.10 + +D LVL: 0.80 → 0.00 → 0.80 +D LPF: 0.20 → 0.90 +D HPF: 0.00 → 0.70 → 0.00 +``` + +These values are normalized musical intentions, not claims about exact acoustic transfer curves. A real capture should preserve the recorded raw points rather than regenerate only these anchors. + +## External MIDI sketch + +C90 is a four-bar D-Dorian polysynth sketch using empty pad C12 on MIDI channel 3. It contains: + +- notes and varied velocities/durations; +- pitch bend into the final chord; +- mod-wheel movement; +- channel pressure after the chord attack; +- one program-change event at the start, recorded/relayed even though current firmware no longer maps it internally. + +C90 is not in the song list but should remain available as a Session clip. The main song uses D12, a resample of its best chord, so sonic playback of the arranged song does not depend on external hardware. A project-aware export should still expose C90's MIDI dependency rather than inventing the missing synthesizer patch. + +## Creation journal: what the musician actually did + +1. Imported drums, bass and Rhodes; recorded the voice and laundromat ambience from line/mic. +2. Assigned groups by role and made several pad variants from the same global slots. +3. Live-chopped the Rhodes phrase across C9–C11, then refined trims without cropping the source. +4. Programmed the core beat with a mix of live recording, step entry, note repeat, free timing and tick nudges. +5. Played B1 and C1 in KEYS mode, using LEG for the bass and KEY for polyphonic Rhodes. +6. Recorded multiple fader assignments into group patterns. +7. Built sections by Commit, then edited the A/B/C/D pattern selection and lengths in each scene. +8. Temporarily selected Filter and performed punch-ins while resampling a two-bar hook. This became slot 620/D11. +9. Sequenced the external polysynth from C12/C90, then resampled the useful chord to slot 621/D12. +10. Selected Delay as the project's final send effect, set output compression and configured note sidechain. +11. Constructed the thirteen-position song list, deliberately repeating and returning to scenes. +12. Kept Scene 10 and C90 as useful experiments even though neither appears in the final song. + +## Live-only performance score + +The musician may perform these gestures while recording a stereo master, but they are not assumed to be part of the project archive: + +- Position 2: momentarily solo Groups A and D. +- Position 5: apply pressure-sensitive beat-repeat punch-in to the final half-bar. +- Position 6: engage the output LOOP buffer, shorten it, slide it backward, then exit on the bar. +- Position 9: combine a tape-stop-style punch-in with a group solo. +- Position 12: manually increase Delay feedback X/Y and release it before the outro. +- Throughout: temporary held-group project-volume changes for performance balance. + +Only the earlier resampled D11 performance is durable. A backup-only export should not fabricate Arrangement automation for this live score. + +## Source gesture versus recoverable result + +| Action | What should survive | What should not be invented | +|---|---|---| +| Live/step pad recording | final source pad, note, velocity, duration and tick | whether it was finger-drummed, step-entered, quantized or nudged | +| Note repeat/arpeggio | resulting note events | original pressure/gesture history unless separately encoded | +| Chopping | per-pad assignment and trim | AUTO/equal/attack/LIVE method | +| Fader recording | raw automation records | a smoothed lane that loses the captured points | +| Scene Commit | resulting scene/pattern references | commit and undo history | +| Song editing | ordered positions including repeats | one-scene-equals-one-position simplification | +| Punch-in FX | nothing assumed as normal pattern automation | imagined punch-in event lanes | +| Resampled punch-in performance | D11 audio and pad state | editable punch-in history | +| Resampled external synth | D12 audio | external patch and oscillator state | +| Live Delay X/Y during resampling | result inside D11 and perhaps final static X/Y | X/Y automation unless real records prove it exists | +| LOOP, FX+Group solo, live scene choices | performance cue only | Arrangement automation unless explicitly stored | +| Silent sidechain trigger | A12 notes and saved routing | audible trigger audio | +| External MIDI sketch | C90 events/channel/root | sound of disconnected hardware | + +## Ableton Live 10 acceptance questions + +1. Does Live open with A–D Group Tracks and separate children for independent pad instances? +2. Does Session View contain Scenes 01–10, including out-of-song Dub Lab? +3. Does Arrangement follow all thirteen song positions rather than unique numeric scenes? +4. Is Arrangement exactly 216 quarter notes long? +5. Do meter markers change 4/4 → 3/4 → 4/4 → 2/4 → 4/4 without shifting later clips? +6. Do short group patterns loop until the longest pattern in their scene finishes? +7. Are repeated song positions identical, including pattern automation? +8. Do A1/A2, C1/C2 and C9–C11 share WAV files while retaining independent pad settings? +9. Does C2 sound `-0.09` semitone detuned rather than roughly `-1.9`? +10. Do all EP pads use non-looping transport, with KEY/LEG behavior represented without Simpler Classic loops? +11. Does A5 choke A6 across separate child tracks? +12. Does B1 continue its playhead through LEG pitch changes? +13. Can C1 play overlapping chord voices from one pad source? +14. Are note velocity, duration, microtiming and pitch variations intact? +15. Are BPM stretch, BAR stretch and reverse distinct, including BAR in the 3/4 scene? +16. Do lower-rate and stereo sources retain intended pitch, length and channels? +17. Is pad AMP independent from group LVL and the runtime-only project volume? +18. Is every fader lane visible, including TIM/MOD when its acoustic mapping is unresolved? +19. Does the selected Delay exist once, with A–D sends and recorded send motion? +20. Are static Delay X/Y distinguished from the baked X/Y performance inside D11? +21. Is output compression after dry groups plus Delay return? +22. Do audible A1 and silent A12 both trigger ducking of B/C/D? +23. Do D11/D12 play as literal rendered audio without invented source automation? +24. Does C90 export as a MIDI-only Session clip with channel/expression and remain out of Arrangement? +25. Is the musician warned about live-only performance cues that no backup can reconstruct? + +## Listening oracle + +- Mute A2, then A1: their roles must differ strongly despite sharing a file. +- Solo A5/A6: closed hat must terminate the open hat. +- Solo B1: the LEG phrase should change pitch without restarting its attack. +- Solo C1: its chord must remain polyphonic. +- Solo C1+C2: C2 should create subtle detuning, not a large downward transpose. +- Mute A1 audio but keep A12 notes: bass, harmony and vocal groups must still duck. +- Compare Scene 03 at positions 3, 4 and 8: it should be identical each time. +- Listen across Scenes 05 and 07: meter changes must not displace later clips. +- Solo D11: the performed effect is already in the WAV and should need no generated punch-in automation. +- Disable the Live master dynamics counterpart: the difference should demonstrate why AMP cannot be judged without output-stage context. + +## Explicit uncertainties + +- The exact acoustic laws for AMP, every fader function, output compressor and sidechain remain measurement questions. +- TIM, TUNE and MOD are legitimate musician controls, but their best Live counterparts are not established. +- Master FX X/Y should not be considered recorded automation without an identified event stream. +- The exact meaning of MIDI-looking project records must be captured before assigning all C90 expression events to binary kinds. +- Per-scene meter is structurally evidenced, but a real mixed-meter project should verify all proposed transitions. +- Human scene/pattern names here are annotations, not assumed archive fields. + +Sources: [official workflow](https://teenage.engineering/guides/ep-133/workflow), [sound and song modes](https://teenage.engineering/guides/ep-133/modes), [recording and fader automation](https://teenage.engineering/guides/ep-133/play-and-record), [sampling/resampling/timing](https://teenage.engineering/guides/ep-133/functions), [effects/output/sidechain](https://teenage.engineering/guides/ep-133/effects), and [current firmware features](https://teenage.engineering/guides/ep-133/whats-new). diff --git a/docs/research/ep133-musician-audit/README.md b/docs/research/ep133-musician-audit/README.md index 96b01ee..36b0d56 100644 --- a/docs/research/ep133-musician-audit/README.md +++ b/docs/research/ep133-musician-audit/README.md @@ -9,7 +9,7 @@ It deliberately does not propose code changes. It supplies a musical reference p ## Files - `MUSICIAN_WORKFLOW_INVENTORY.md` describes how a musician can build and perform an EP-133 project and separates saved project state from rendered audio and live-only gestures. -- `IMAGINARY_PROJECT_AFTERIMAGE_TRANSIT.md` specifies one coherent imaginary project in enough detail to act as a translation oracle. +- `IMAGINARY_PROJECT_NEON_LAUNDROMAT.md` specifies one coherent imaginary project in enough detail to act as a translation oracle. - `LIVE10_TRANSLATION_AUDIT.md` asks whether the current group-track Live 10 export represents every decision in that imaginary project. ## The three kinds of musical truth