This is the REAL, currently-flashed 5-trial design (replaces the old
family/stability/donor-recipient/repair/Trial-6 design entirely — that design is gone,
not just superseded). Answers and audio track numbers below are read directly out of the
real firmware constants (cross_pollination.ino) and the real delivered audio
mapping (TRIAL_AUDIO_MAPPING.md), not guessed.
AUDIO_SCRIPT.md.
One line per touch. Box numbers are 1-9, front view, top-left = box 1, reading left-to-right/top-to-bottom (matches the lowest-to-highest hall-sensor pin order).
This exact sequence — 1, 7, 3, 4, 8, 4, 6, 3, 1, 4, 6, 8, 9, 2, 6, 7, 9 — is a real, complete, valid playthrough from idle to finale.
Unit 6's clues were rewritten in October to match what is really in its
boxes (photos: boxes/), so the answers differ. Same five trials and step types.
Full playthrough: 5, 6, 3, 4, 8, 8, 7, 3, 9, 4, 6, 8, 5, 2, 1, 7, 9.
Lighting rules (1.10.1): blacklights stay on. During the trials every box has its own color, and a box nobody has picked is off. A picked box lights in its own color (a held pick steady, a confirmed answer with a gentle shimmer); a wrong touch flashes red. When the narration refers to a box it blinks in its own color, timed to the words (e.g. "Return to the three broadleaf specimens"); after a "look at these" blink the boxes stay lit at 75% until the next correct answer. A correct answer flashes only the correct boxes, square and blacklight together. After a box is picked it ignores the wand for 1.5 s and until the wand has been lifted off it for a second, so a wand left resting on a box doesn't pick it again.
Idle (attract): blacklights on, every square glowing its own color with the odd tube-style flicker. The first wand touch on any box starts play and doesn't count as an answer. The squares flicker on one at a time in twitching random colors, like fluorescent tubes striking, hold briefly, then go dark. The Trial 1 opening plays.
The walkthrough below uses the standard cabinets' box numbers. On unit 6 the lighting is the same but the boxes follow the unit 6 key above.
Touch box 1 and box 7, in either order.
✓ First correct touch: acknowledgment cue, box holds lit in its own color (re-touching it is a no-op). Second (the other member): both boxes flash gold together, then stay lit, pair confirmed, 1 and 7 flash (square + blacklight), advance to Trial 2. Every other square goes dark at this handoff.
✗ Wrong box: that box flashes red then reverts + wrong-answer cue. A wrong SECOND touch does not clear your already-correct first pick.
Touch boxes 3, 4, and 8 — any order, all three required. Distinct cues for 1st/2nd/3rd correct touch.
✓ Each correct touch stays lit immediately (no held pick in this trial). Once all 3 are found ("you identified all three broadleaf specimens"): every square goes dark and 3/4/8 flash (square + blacklight). When the Trial 3 opening says "Return to the three broadleaf specimens", 3/4/8 blink again and stay lit until the next correct answer.
✗ Wrong box: that box flashes red then reverts; "incorrect first selection" cue before you've found any, "incorrect retry" cue after you've found at least one.
Touch box 4 (Stage A), then 6 (Stage B), in that order. 3/4/8 blink as the opening line starts. Stage B correct ends the trial with "Correct... Trial 3 complete" (04/005).
✓ Each stage's correct box stays lit on touch. Stage A correct ("Now examine the entire cabinet") blinks the whole cabinet. After Stage B ("you identified the unique colour"): squares go dark and 4 + 6 flash. When the Trial 4 opening says "Look again at the three broad leaves", 3/4/8 blink and stay lit. On "Now examine the other specimens" the other six blink and then stay lit with the three while the narrator talks; on "First, find the broad leaf" the six go off and only 3/4/8 stay lit, because the first pick must be one of them.
✗ Wrong box at any stage: that box flashes red then reverts + that stage's own incorrect cue, stay on the same stage.
⚠ Stage C (box 8, "the colour that appears only once") was dropped on 2026-09-27 and removed from the firmware in 1.9.0: its opening and incorrect lines were never recorded.
Three steps, each a donor touch then a recipient touch:
✓ Correct donor: held lit, awaiting that step's recipient. Correct recipient: both boxes flash violet together — a distinct "you created a hybrid" look, not the plain flash every other correct answer uses — then that pair flashes (square + blacklight) and stays lit. After cross 1, when the line reaches "Of the two unused broadleaf colours", 4 and 8 blink and stay lit until you pick. After the 3rd cross ("All three combinations are stable"), every square goes dark and all six cross boxes (1, 3, 4, 6, 8, 9) flash as Trial 5 begins.
✗ Wrong donor or wrong recipient: that box flashes red then reverts + that step's own incorrect cue; a wrong recipient touch does not clear the already-confirmed donor.
Touch 2 → 6 → 7 → 9, in that exact order.
✓ Each correct touch lights immediately and stays lit as you move to the next step — nothing resets between these 4 steps, they accumulate. After step 2 ("Keep both specimens active"), and again when "Now compare your two selected specimens" starts, 2 and 6 blink. After step 3 ("three specimens are now active"), 2/6/7 blink. Steps 1-3 each have their own "after correct" cue; steps 2 and 3 also queue a follow-up clue cue a few seconds later. The 4th (final) correct touch triggers the finale lights, the Trial 5 completion cue, then a Hybrid Index ending.
✗ Wrong box at any step: that box flashes red then reverts + that step's own incorrect cue.
All 9 boxes glide slowly through colors for the whole finale audio. One of 3 real Hybrid Index endings plays, chosen at random (1/2/3 — NOT 1-6 like the old design). Per the boss's locked doc, the Hybrid Index sends players to an off-cabinet examination table + Field Guide — no physical hybrid is built in the cabinet itself. When the ending finishes, two white flashes, then the cabinet returns to idle by itself.
Real folders/tracks as laid out on the official show SD
cards (register: platform/tools/cross_pollination/official-sd-cards.md). Unit 6
uses its own card (audio set cross-pollination-sd-u6, u6-2026-10-09-lavender)
with its re-voiced clues in the same folders.
| Folder | Trial | Confidence |
|---|---|---|
| 01 | "Box 1" … "Box 9" (box test only) | On every official card; a card without it is an old-layout card and the box test is silent |
| 02 | Trial 1 | High |
| 03 | Trial 2 | High |
| 04 | Trial 3 | Transcribed; Stage C lines never recorded (Stage C removed), and 04/006 is actually a Trial 4 line |
| 05 | Trial 4 (13 tracks) | Transcribed; firmware remapped 2026-09-27 (see AUDIO_SCRIPT.md) |
| 06 | Trial 5 | High |
| 07 | Hybrid Index endings (3 tracks) | High |
| 08 | Welcome (40.0s) | Present, not wired into gameplay yet — trigger point (attract-loop vs. Mega-triggered one-shot) still needs a decision |
| 10 | Idle warning ("will reset in 30 seconds") | Placeholder voice until the real recording exists |
Full script (what each clip actually says):
platform/tools/mega_dumb_slave/audio/sd-card-root/AUDIO_SCRIPT.md.
Full per-track filename mapping:
platform/tools/mega_dumb_slave/audio/sd-card-root/TRIAL_AUDIO_MAPPING.md
in the repo.
Box numbering: front view, top-left = box 1, reading left-to-right then top-to-bottom — box 1 is always the LOWEST pin number.
| Box | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|
| Mega pin | 22 | 24 | 26 | 28 | 30 | 32 | 34 | 36 | 38 |
Mega pin 12 to GND. Tap = reset the game; hold 2–6 s and release = box test on/off; hold 6 s = photo mode (every square white, blacklights on, game frozen; tap to leave).
| DFPlayer pin | Connects to |
|---|---|
| VCC | Mega 5V |
| GND | Mega GND |
| RX | Mega pin 16 (TX2) |
| TX | Mega pin 17 (RX2) |
Firmware talks to the DFPlayer on the Mega's hardware Serial2 at 9600 baud — needs a Mega (or other multi-UART board), not a plain Uno/Nano.
Real, bench-confirmed: DFPlayer's line-level audio output (DAC_R/DAC_L or R/L, not SPK_1/SPK_2 if the speaker is powered/amplified) → jack pin 2 only. No separate ground wire needed — the speaker gets ground through its own USB power connection. Confirmed working live, 2026-09-25.