A Salaried Illustrator's pixivFANBOX: ¥2M Over 4.5 Years — Down From a ¥40K Peak to ¥25K a Month, Still Running

Revenue disclosures from creative work are usually written while the line is still going up. What makes this one unusual is that it summarizes four and a half years from a point past the peak.

In May 2026, an illustrator who works a salaried job published a retrospective note on running a pixivFANBOX. Cumulative earnings from FANBOX alone come to roughly ¥2 million. But current monthly revenue is ¥25,000–27,000, down from a period when it approached ¥40,000. As the author puts it: starting a FANBOX is easy. The hard part is continuing.

In absolute terms this sits at the small end of the cases. It matters as a record because one person has published the trajectory of one activity across four and a half years, including months where the breakdown is itemized. What you get here (how it declines, and how it keeps going) does not appear in one-off success reports.

The disclosed figures

A finer breakdown appears in the monthly disclosure for May 2023.

Fifty-six supporters in total. Multiply out the plan tiers and it lands exactly on ¥39,500. More than 70% of supporters are on the cheapest ¥500 tier, and that tier also accounts for a bit over half the money. Structurally, this means adding people works better than raising the price.

The same post reports that one supporter who had been on the ¥5,000 plan in April moved down to ¥500, taking the top tier from two people to one. At a scale of 56 supporters, one person moving between tiers shifts about a tenth of revenue. The author’s response is that keeping normal operations during periods when supporters leave matters more than the fluctuation itself.

One more caution on reading the ¥2 million cumulative figure: there were hiatus periods across those 4.5 years, so it is the sum of a line that rose and fell between the near-¥40,000 peak and today’s ¥25,000–27,000, not the result of a steady monthly inflow. It is not a number you can average into “X per month.”

What got dropped after the peak

The most concrete part of this record is not a list of growth tactics but a list of things stopped.

Around 2023–2024 the author gave up on maximizing technical skill. The reasoning: invested hours don’t scale into revenue, and chasing the fine detail work that stretches production time per piece breaks the posting cadence instead. Running solo, an unsustainable schedule ultimately destroys momentum.

What replaced it: a style they personally like, characters and a world they enjoy, and a posting frequency they can sustain. The design shifted from defending a quality ceiling to defending a frequency floor.

Generative AI is the backdrop. Since NovelAI’s arrival in October 2022, image generation has advanced to where technical execution is hard to distinguish from human work. The author accepts that, and argues that the motive of drawing because you love it, and years of stubborn devotion to one genre, is harder to replace. On AI’s effect on stock-asset markets there are records like a stock photographer’s pivot after AI and selling AI-generated images on Adobe Stock. This case runs the other direction: moving the product toward ground where it does not compete with AI.

How to read the amount

¥25,000 a month is small as a side income for a salaried worker. But the structure gives it a different character.

The cost of goods sits near zero. FANBOX handles payments and delivery, and what’s left after the platform fee is the net (in May 2023, ¥35,550 on ¥39,500 gross). No inventory, no shipping. Income, on top of that, is not contingent on commissions. Commissioned illustration is one of the areas most exposed to generative AI, but FANBOX support is payment for continuing to want to see this person’s work, which sits outside price competition.

The tier design is worth looking at too. ¥500 / ¥1,000 / ¥5,000 spans a tenfold range, and yet only one person used the top tier, contributing 13% of revenue. Far from being a pillar, the top tier is a welcome exception. Which means there is no path to stability other than thickening the bottom tier. This is the familiar distribution of patronage-style monetization, shown plainly.

And time is doing work here. Even at ¥25,000, fifty-four months adds up. Small-amount-times-long-duration is the same shape as a single note article that kept selling for five years and the first months of a paid subscription magazine. The number is built from months not quit, not from the size of any single hit.

A ceiling of 56 supporters, and a 30–40% slide from peak

The weakness is plain: the ceiling is low. Fifty-six supporters is a function of follower count and posting frequency, and it’s bound by the output someone can sustain alongside a full-time job. The ¥500 tier at 70% is evidence of a low barrier to entry, but it also shows that average revenue per supporter barely moves. Changing the order of magnitude means multiplying supporter count several times over, which is hard to do while employed.

The decline of roughly 30–40% from peak also deserves to be handled head-on. There were hiatus periods. Generative AI changed the environment. The author deliberately reduced detail work. How much each contributed is not disclosed. Without that decomposition, the decline cannot be read as a rule of any kind.

The starting point was small, too. The record notes roughly 400 pixiv followers at the outset, and rather than aiming at a large monthly figure immediately, the author recommends stepping through ¥5,000, then ¥10,000. This is not a story that began from a large base.

And the element the author considers most important, the persistence of drawing the same genre for years, is the one hardest to measure from outside. More than the size of the figures, the value of this record is that four and a half years of them were published at all.