G: 音楽（おんがく）箱（ばこ）を作（つく）る — Make a music box.
いちばんかんたんな方法（ほうほう）は紙（かみ）テープ型（がた） — The easiest method is the paper-strip type.

Most reliable path for a “tech-adept generalist”

Option A (most reliable + least custom machining): paper-strip DIY music box (15/20/30-note)

This is the sweet spot because you’re not machining a pinned cylinder; you’re just generating a punch pattern.

Workflow

Pick a mechanism format: 15/20/30-note paper-strip crank music box.
- 30-note gives you the best chance of keeping “Quentin’s Theme” recognizable without brutal simplification.
Get the melody into MIDI (or enter notes manually).
Convert/arrange for the note range using Music Box Maniacs’ editor (imports MIDI; exports printable punch templates). citeturn0search0turn0search10
Export a PDF punch pattern, then:
- Hand-punch (slow but simple), or
- Laser cut (more accurate and repeatable).
Test → revise: you’ll almost always need to re-voice chords, drop inner notes, or shift octaves to fit the limited note set.

Making it “reliably clean” (laser instead of hand punching)

A ready path is generating vector files from MIDI and cutting them.
- Example: the MBM community “MIDI → DXF” approach for laser cutting strips. citeturn0search2
- Or use a CLI tool that outputs SVG punch strips from MIDI (good if you like reproducible pipelines). citeturn0search9

Option B (mechanical wind-up movement): buy a Sankyo movement and build the box

If you want a classic wind-up feel, you can buy an 18/23/30-note movement module and build the enclosure. citeturn0search3turn0search5turn0search7
But: these movements are usually pre-pinned for one song. Custom pinning for a specific tune is the hard part, so this option is great for “custom box,” not great for “custom song.”

Option C (best playback fidelity, but not “classic”): programmable mechatronic music box

If you want the tune to sound right and don’t care that it’s electromechanical under the hood, a programmable box that plays from MIDI is the “it just works” route (higher cost, lowest frustration). One example is Muro Box (aimed at MIDI arranging workflows). citeturn0search18turn0search11

Getting Quentin’s Theme into your pipeline

The tune you mean is commonly marketed as “Quentin’s Theme (Dark Shadows)” and is associated with Bob Cobert (often listed as Cobert/Grean in some publications). You can still start from ear/MIDI, but sheet music helps. citeturn0search12turn0search8

Practical “do this, not that”

Do 30-note paper strip if you want it recognizable without becoming a nursery rhyme.
Do laser-cut strips if you care about timing accuracy and repeatability. citeturn0search2turn0search9
Don’t attempt a custom pinned cylinder unless you want a machining project; it’s the least reliable path for a generalist.

Third option / omission: legality/release reality

If you plan to sell the resulting strips/boxes with “Quentin’s Theme,” you’re in copyright/licensing territory. For personal use, it’s usually fine; for distribution, assume you need permission/licensing, especially if you’re copying from published sheet music. citeturn0search12turn0search8

Baseline reality check: how long is “Quentin’s Theme,” vs. a typical cylinder music box

“Quentin’s Theme” (Charles Randolph Grean Sounde version) is listed at ~1:59. citeturn1search5
A very common 18-note pinned cylinder movement gives you ~14–18 seconds per one full cylinder revolution (then it repeats that same “one-revolution song”). citeturn0search3turn0search1
Rule of thumb: depending on comb/note count, one revolution is often ~12 seconds up to ~50 seconds. citeturn0search2
A (bigger) 36-note cylinder scale is quoted as ~25–32 seconds of melody length per tune. citeturn0search0

Comparison in plain terms

If you try to include the whole ~1:59 (≈119 seconds) tune at normal tempo:
- vs 17 seconds/rev (typical 18-note): you need about 7 revolutions worth of “time axis.”
- vs 30 seconds/rev (bigger cylinder/comb class): you still need about 4 revolutions worth.

A normal cylinder music box design doesn’t give you 4–7 revolutions of unique pattern; it gives you one revolution and then it loops.

What “machining a longer tune” really means (mechanically)

A pinned-cylinder music box encodes time around the cylinder’s circumference. The comb’s teeth are the note set; the cylinder’s pins are the “MIDI timeline.”

So to get a longer, non-looping tune, you must change at least one of these:

1) Make the cylinder’s “time axis” longer than one revolution

A. Bigger circumference cylinder

Keep the same tempo and pin spacing → you need a cylinder with ~4–7× the circumference of a typical movement to hold the whole ~2 minutes in one pass.
That quickly becomes physically huge (and demands more torque, stronger governor control, and stiffer alignment).

B. Helical / shifting-cylinder design (industrial-grade approach)
Instead of one ring of pins repeating, the cylinder moves laterally so the comb encounters a new band of pins as it rotates.

Reuge describes “changeantes” where the cylinder shifts after a full rotation to engage a new set of pins for another melody. citeturn2search22
In principle, you can extend that idea from “shift once per tune” into a continuous helical advance (like a leadscrew) so you get a long, single program.
This is the real way to do “long cylinder program” without making a dinner-plate cylinder, but it’s a precision mechanism project.

C. Disc music box (often the practical “long song” mechanical format)
A disc can pack more “timeline” area without insane diameter, because you can encode radially/spirally. It’s why disc mechanisms historically handle longer pieces more comfortably than small pinned cylinders (still not “any song,” but longer is more plausible).

2) Increase note count (helps musicality; only partly helps length)

More notes = more faithful melody/harmony, but doesn’t magically make the timeline 7× longer. It can push per-revolution tune length upward (12→50 sec range is commonly cited), but you’re still fighting the loop. citeturn0search2

3) Slow the cylinder down (tempting, but it changes the music)

You can gear it slower so one revolution takes longer, but:

if you don’t change the pin pattern density, the melody slows down and feels wrong;
if you increase pin density to compensate, you hit mechanical limits (pin spacing, missed plucks, damping conflicts).

What you’d physically have to fabricate / modify (the checklist)

Assuming you mean a traditional pinned cylinder + comb (not paper strip, not stepper motors):

Core parts you must control

Comb (the tuned “keyboard”)
- Number of teeth = available notes (e.g., 18/30/36/72…). citeturn2search22turn0search0
- Tuning is non-trivial; buying a comb/movement is easier than making one.
Cylinder (the song data storage)
You need:
- Accurate diameter, runout control, and axial alignment so pins hit the correct teeth every time.
- A precise indexing grid for where pins may go. (DIY builds often lay out the circumference into equal divisions and drill a matrix of candidate holes.) citeturn0search14turn2search12
Pins (and their geometry)
- Consistent pin height, shape, and retention (press-fit / peened / soldered, depending on design).
- The pin must lift/release the tine cleanly without snagging neighboring teeth.
Governor + spring motor + gear train
- Keeps tempo stable across spring unwind.
- Any “long tune” approach that increases cylinder size or adds lateral motion increases torque needs and instability risk.

The “long tune” add-ons (where it becomes a real engineering project)

A lateral shift mechanism (if you want >1-revolution unique playback without a gigantic cylinder)
- A cam/ratchet/lead screw that advances the cylinder (or the comb) predictably.
- Return/reset at the end so it doesn’t just walk off the rails.
Drag management (polyphony limits) If you try to pluck multiple notes at once too often, friction/drag rises and the mechanism can bog down. One vendor explicitly recommends arranging so most of the song is one note at a time, with two-note moments and maybe a final chord, because of drag. citeturn0search3

So what does this mean specifically for Quentin’s Theme (~1:59)

A typical 18-note “custom cylinder” service targets something like ~8 bars / ~17 seconds and then loops. citeturn0search3turn0search1
The full ~1:59 piece is **~7× that length**, so if you insist on “traditional cylinder plays the whole thing once,” you are basically choosing between:
1. Very large custom cylinder (physically big, torque-heavy, hard to stabilize), or
2. A cylinder that shifts laterally over multiple bands (precision mechanism; the “right” mechanical solution), or
3. Disc format (often the sane mechanical way to get longer playback area), or
4. Accepting a curated excerpt (the classic music-box reality) and picking the most iconic ~15–30 seconds.

If what you want is “a longer tune, still purely mechanical, still traditional,” the most reliable engineering path is the lateral-advance cylinder concept (a controlled helix / multi-band program), because it scales length without comically scaling diameter, and it matches how high-end makers extend capability (even if they usually shift per tune rather than continuously). citeturn2search22turn0search10