Geom
This site is a sketchbook and working log for creating a music video for Musical Architecture by Yaporigami (Yu Miyashita). It captures the thinking and behind-the-scenes process, with the log itself serving as an introduction to Sketching with Math and Quasi Physics.
So we talked about making the video very geometric. Setting constraints and grounding everything in strict logic felt like a fun project—especially in a time when you can ask for anything (to AI) and get it without knowing how it’s made. But where should we draw the line?
Polygons and Bézier curves are mathematically defined, and almost any shape can be decomposed into sine waves via Fourier transform, so in theory we could claim that anything can be described mathematically.
We set a set of tentative, self-imposed rules.
- We rely exclusively on deterministic, fundamental functions (e.g., polynomials and trigonometry) and their iterations, extended into complex numbers, while avoiding any manual crafting of geometry.
決定論的で基本的な関数(多項式や三角関数など)とその反復、複素数への拡張のみに依拠し、ジオメトリを手作業で作り込むことは避けます。 - We keep the visual generation process completely deterministic: the same input (sound data) produces exactly the same result. We also avoid random and noise functions (even though computer “randomness” is pseudo-random and therefore deterministic). Sound data is the only quasi-random element, in the sense that it yields varying values without following the mathematical functions described above.
映像生成のプロセスは完全に決定論的に保ち、同じ入力(音声データ)からは常に同じ結果が得られるようにします。乱数関数やノイズ関数も避けます(コンピュータの「乱数」は疑似乱数であり、本来は決定論的ですが)。また、音声データだけは、上記の数学関数に従わずに変化する値を返すという意味で、唯一の疑似ランダムな要素となります。 - We minimize arbitrary choices. This means avoiding keyframing whenever we can. It turns out it is nearly impossible to follow this rule strictly, since any choice—such as which function to use and when, and how we process sound into data—is itself arbitrary. Still, we try to drive changes, transitions, and other variations through simple, deterministic rules as much as possible.
恣意的な選択を最小限にします。つまり、可能な限りキーフレーム(手打ちのアニメーション)を避けるということです。もっとも、このルールを厳密に守るのはほぼ不可能だと分かってきました。どの関数をいつ使うか、音声データをどう処理するかという選択自体が、すでに恣意的だからです。それでも、変化や遷移、その他のバリエーションはできる限りシンプルで決定論的なルールによって駆動させるよう努めています。