We first heard about this project the way we hear about most good ones: a reader emailed us a photo. A subscription-member in Portland had received a limited vinyl box set — 300 hand-numbered units, ambient piano, pressed at a small plant in the Netherlands — and the cover looked like no two copies were alike. They weren't. The label had commissioned Tokyo-based artist Jun Fujiki to design a generative system that produced 300 unique covers from a single set of rules. We followed the project for four months, interviewed the label's two founders (who asked to stay pseudonymous), and tracked the numbers. Here's what happened.
The Brief and the Constraint
The label's original plan was simple: one cover, one print run, one SKU. But the pressing plant quoted a minimum of 300 units, and the founders worried that a single static cover would sit invisibly on shelves. They wanted each copy to feel like an object, not a product. The constraint was brutal — no budget for 300 bespoke paintings, no time for a traditional illustration cycle, and a deadline of eleven weeks before the pre-order window opened.
That's when a designer on their team suggested generative art. The idea: define a visual system — geometry, folding logic, a constrained color palette — and let an algorithm output 300 distinct compositions. The label reached out to Jun Fujiki, whose portfolio spans generative art, paper-folded installations, and computational design, with a question that was really a test: could a system built for gallery-scale work survive the indignities of a 12-inch square, 300 times over?
Timeline: Eleven Weeks, Four Decision Points
Week 1–2 — System design. The artist proposed a folded-paper logic: each cover would simulate a single sheet creased along one of several axes, then lit from an angle. The palette was locked to three inks plus black. The label pushed back on the palette — too austere for a debut release — and the artist added a fourth accent color drawn from the album's mood board. This was decision point one, and it cost four days.
Week 3–5 — Iteration. The first 50 test outputs were, in the label's words, "beautiful but samey." The folding logic was too consistent. The artist introduced controlled randomness: rotation variance, crease-depth variance, and a probability that a fold would fail and be redrawn. Decision point two: how much chaos? Too little and the covers read as a pattern; too much and they read as noise. They settled on a variance band that produced roughly 12 distinct visual families across the run.
Week 6–7 — Print proofing. This is where most generative projects die. Screen colors and ink colors are not the same species. The label ran three proof rounds; the first two failed on the accent color, which shifted from a warm terracotta to a muddy brown on uncoated stock. Decision point three: switch to coated stock or change the accent. They changed the accent.
Week 8–11 — Production and numbering. The final files were exported as 300 individual PDFs, each with a unique identifier embedded in the artwork itself — a small geometric mark in the lower right that encoded the unit number. Decision point four: should the numbering be visible or hidden? They chose hidden, and the result is that collectors have to look for it. That small choice turned the box set into a puzzle, and puzzles get posted online.
Obstacles We Didn't Expect
Two problems emerged that no one forecasted. First, file management: 300 unique print files meant 300 opportunities for a naming error, and one mislabeled file shipped with the wrong unit number. The label caught it before fulfillment but only because a volunteer checked all 300 by hand. Second, and more interesting, the pre-order page needed to show the covers — but which ones? Showing 300 images was impossible. The solution was to show 12, one from each visual family, and let buyers choose a family rather than a specific unit.
That decision changed the marketing. Instead of "buy the record," the pitch became "pick your fold." Pre-orders opened on a Tuesday and sold out in 41 hours. The label had projected a six-week sell-through.
Measurable Results
- Sell-through: 300/300 units in 41 hours, versus a six-week projection.
- Engagement: The label's Instagram post showing 12 cover variants outperformed its previous best post by 4.9x on saves.
- Secondary market: Within a month, individual copies were reselling at roughly 2.3x face value, with buyers specifically hunting for the rarest fold family.
- Cost: The generative commission plus proofing came in under the quoted cost of a single bespoke illustration plus licensing.
The takeaway for anyone running a vinyl subscription or a small pressing: uniqueness is a feature you can manufacture, and it doesn't require manufacturing 300 paintings. It requires a system, a tight brief, and a willingness to let go of some control. The label's founders told us they'd do it again — but with a stricter file-naming protocol and a dedicated proofing budget.
If you want to see the underlying logic — the fold families, the geometry, the way a single crease becomes a variable — the artist's process notes are worth an afternoon. Start with the generative art and computational design work on the portfolio, then work backward through the paper-folded installations. The box set is a small commercial application of a much larger practice, and seeing both sides makes the 300 covers feel less like a gimmick and more like a translation.