Bytes of Art ■ Episode 01
4,096 bytes. That is less than this paragraph. Less than most favicons. Less than an empty Word document. It is, to put it precisely, four kilobytes.
Zetsubo by Prismbeings (2018) is a complete audiovisual experience compressed into that budget: procedural geometry, an atmospheric soundtrack, real-time rendering. No textures. No samples. No assets. Everything generated from mathematics at runtime.
The name comes from 絶望 (zetsubou), Japanese for "despair." The mood matches: melancholic synths, slow-building atmosphere, a dying world rendered in shades of grey. One does not watch Zetsubo so much as inhabit it for two minutes, then spend the next twenty wondering how on earth it was done.
Watch it. Then come back. The rest will make more sense.
The Size of Nothing
Four kilobytes is a number that means very little until you place
it next to other numbers. A favicon, the tiny icon in your browser
tab, typically weighs 15 KB. An empty .docx file
created by Microsoft Word: 12 KB. The minified production build of
React: 136 KB. jQuery, a library many considered "lightweight":
87 KB. A single hero image on a modern landing page: 200 to 500 KB.
Zetsubo fits inside all of them. Comfortably. With room to spare for a second Zetsubo.
The gap is not a rounding error. It is a philosophical chasm.
On one side: an entire world, with sound, light, and motion.
On the other: a library that helps you write
document.getElementById() with fewer characters.
How Four Kilobytes Become a World
Zetsubo contains no pre-made content. No image files, no audio samples, no 3D models, no texture maps. Everything you see and hear is generated from mathematical formulae the moment the executable runs. The rendering pipeline, if one can call it that at this scale, works roughly as follows.
Geometry: raymarching with Signed Distance Functions.
Traditional 3D rendering uses meshes: thousands of triangles
describing shapes. Storing triangles costs bytes. Zetsubo uses
Signed Distance Functions
(SDFs) instead. An SDF is a mathematical function that, given any
point in space, returns how far that point is from the nearest
surface. A sphere is one line of code: length(p) - radius.
Complex shapes are built by combining simple SDFs: unions,
intersections, smooth blends. The geometry exists as equations,
not data.
Rendering: raymarching. For each pixel on screen, a ray is cast from the camera into the scene. The renderer calls the SDF, takes a step equal to the returned distance, and repeats. When the distance becomes negligibly small, the ray has hit a surface. From that hit point, the renderer calculates normals (for lighting), occlusion (for shadow), and atmospheric scattering (for mood). No triangle rasteriser. No Z-buffer. Just a function, evaluated millions of times per frame.
Sound: a synthesiser written in code. There is no audio file. The soundtrack is generated by a software synthesiser built into the executable: oscillators, envelopes, filters, all expressed as arithmetic operations. Note by note, sample by sample. The melancholic pads and the slow rhythmic pulse are not recordings; they are equations that happen to sound like despair.
Visuals: shaders simulating physics. Light, shadow, fog, and atmospheric perspective are all computed per pixel in real time. The grey, desolate palette is not an artistic shortcut to save bytes (though it does save bytes); it is the emotional register of the piece. The constraint and the aesthetic are the same thing.
The Culture Behind It
This is demoscene. A subculture that has existed since the early 1980s, originating in the cracktro scene: the signature animations that software pirates attached to pirated games. "We cracked this. Here is our calling card." Those calling cards became the art form. The pirates became programmers, musicians, and graphic artists. The signatures became standalone productions, judged at competitions called demoparties.
In 2020, the demoscene was inscribed on the UNESCO Representative List of the Intangible Cultural Heritage in Germany, Finland, and Poland. Hardly anyone noticed. A global art movement that has been proving for four decades that constraints foster creativity, recognised as cultural heritage by the same body that protects Venetian gondola craftsmanship and Japanese Noh theatre. The irony is that the demoscene would have preferred a smaller certificate.
The competition categories tell the story. A "4K intro" must fit in 4,096 bytes. A "64K intro" in 65,536 bytes. A "demo" has no size limit but is judged on artistry, technical achievement, and design. The constraints are entirely artificial. Nobody forces anyone to work within 4 KB. They choose to. And within those constraints, they produce work that engineers with unlimited budgets and unlimited dependencies rarely match: coherence, intention, and the quiet confidence that comes from understanding every single byte in the binary.
The Question That Lingers
While we debate whether 500 MB JavaScript bundles are acceptable, a subculture has spent forty years answering a different question: "How much beauty fits in four kilobytes?"
No frameworks. No dependencies. No npm install.
No build pipeline that takes longer than the piece itself.
Just mathematics, a compiler, and the discipline to make every
byte justify its existence.
The answer, as Zetsubo demonstrates rather convincingly, is: more than you would believe. A world rendered in shades of grey, with a soundtrack that earns its name. Four kilobytes of despair, and every single one of them earned.
The question is never whether constraints limit creativity. The question is whether unlimited resources have ever produced it.