Bad Apple!! — A Stage Theme That Became a Benchmark

There is a short, high-contrast animation of black silhouettes flowing across a white field, set to a vocal remix of a video-game tune, that has been made to play on an oscilloscope, a pocket calculator, a smartwatch, a chip costing pennies, a character LCD with no processor behind it at all, a breadboard full of logic chips with no CPU, and — with a certain amount of malice — inside a font file. Its name is “Bad Apple!!”, and how a stage theme from a 1998 doujin shooting game became the informal proof-of-life test for improbable hardware is a story with three distinct stages.

1.1 The tune

“Bad Apple!!” began as background music. It is the stage three theme of Lotus Land Story, the fourth game in the Touhou Project series, released in 1998 and written — music, code, art and all — by the developer known as ZUN under the Team Shanghai Alice name. The original is a chiptune, composed with frequency-modulation synthesis for the NEC PC-9800 family of Japanese personal computers, at 161 beats per minute. It is a good tune, but on its own it would be one of several hundred pieces of Touhou music with a devoted following and no wider reach.

Touhou’s licensing is the reason the story continues. The series has long permitted, and effectively encouraged, derivative work, and a substantial arrangement scene grew up around it. A remade version of the track appeared on Akyu’s Untouched Score Volume 1 on 21 May 2006.

1.2 The remix

The version almost everyone has heard is not ZUN’s. It is a vocal arrangement by the doujin circle Alstroemeria Records, sung by nomico, released on the album Lovelight. Sources consulted for this dive disagree on the date — most give 2007, at least one gives May 2008. The MusicBrainz release record linked above dates the album to 20 May 2007 (Alstroemeria Records, catalogue ARCD0018); that database is community-maintained, so it is offered as the best available evidence rather than as a settled fact. The arrangement turns a driving instrumental loop into a pop song with a hook, which matters, because the next stage required something people would want to watch to the end.

1.3 The video

On 26 October 2009, a Niconico user going by Anira uploaded a music video for a shortened cut of the Alstroemeria arrangement. It was made in response to a request from another user, who had posted a rough storyboard version a year earlier and asked for someone to animate it properly.

What Anira produced is shadow art. Touhou characters — Reimu Hakurei, Marisa Kirisame and others — appear as flat black silhouettes against white, dancing and flowing, with each figure dissolving into the next through negative space rather than cutting. There is no line work, no shading, and no colour. It is a rotoscoped-feeling piece built entirely out of filled shapes.

It went further than anything in the scene had. It reached number one on Niconico’s daily rankings on 15 November 2009, and it became the first video on the platform to pass ten million views — a milestone that mattered on a site where view counts were the public scoreboard. By late 2025 the Niconico upload had passed thirty-one million views, and reuploads carried it to audiences that had never played a Touhou game. In 2020, Alstroemeria Records recognised Anira’s animation as the official music video for their own track, which is an unusual inversion: the fan work became canonical.

A video account of the animation and its origins is also available.

1.4 The benchmark

The third stage is the one this dive is about, and it has no single origin date. At some point the animation stopped being a thing people watched and became a thing people ported.

The usual comparison is to Doom, which has been made to run on printers, ATMs, pregnancy tests and a great deal else, to the point that “does it run Doom” is a recognised genre. The comparison is useful but imprecise, and the difference is the interesting part. Doom is an interactive program: porting it demands a CPU capable of running its logic, enough memory for its data structures, and an input path. Getting Doom onto a device proves the device is a computer.

“Bad Apple!!” is a video. Playing it demands no logic at all — only that a device can be made to emit the right pattern at the right time. That makes it a much broader test, and a much shallower one. It reaches hardware that could never host Doom: a memory chip wired straight to a display, a stereo audio output driving deflection plates, a text-shaping engine inside a font. What it proves about each is correspondingly narrower, and volume four takes up that question directly.

One property of the original does all the work. The animation is one bit deep. There is no grey, no anti-aliasing, no gradient. A frame is a set of regions that are either black or white, and the boundaries between them are long and smooth. That is not a compromise made for the sake of portability — it was a stylistic choice made for a fan music video in 2009 — but it is the reason the thing travels. Volume two is about what that single bit buys.

1.5 What this dive covers and what it does not

Volume two works through the three budgets any port has to satisfy — storage, bandwidth and decode time — and the compression techniques that repeatedly appear when people squeeze the animation into hardware that should not hold it. Volume three surveys ports that can be checked against a published writeup, grouped by what kind of display each one drives. Volume four asks what a successful port actually demonstrates, and separates that from what is commonly claimed.

A caveat on sourcing is owed up front. There is no authoritative index of “Bad Apple!!” ports, and there cannot be one: the practice is decentralised, much of it lives in video form, and a great deal of it is undocumented. The ports described in volume three are those with a project writeup, a source repository, or coverage in a publication that examined them. Claims encountered only as a video title or a forum assertion are noted as such and are not presented as established.