//Glyph Generation

▸How Edenji Are Created

Every edenji character is procedurally generated from a seed word using deterministic hashing. Given the same seed, the engine always produces the same glyph — enabling caching and reproducibility across the application.

▸Algorithm Overview

Step 1: Seed Hashing

The input word is hashed to produce a deterministic pseudo-random number sequence. This seed drives all subsequent decisions: radical selection, layout choice, and character type.

seed("aine") → hash → PRNG sequence

Step 2: Character Type Selection

The hash determines which of six character types to generate, weighted by frequency:

Step 3: Radical Selection

Inspired by the character type, radicals are selected from the 256-slot radical library:

• ▪Semantic radicals are chosen by meaning — the hash maps the seed's semantic domain to matching radicals
• ▪Phonetic radicals are chosen by sound — the hash maps the seed's pronunciation to radicals with similar readings
• ▪The number of radicals (1–3) depends on the character type and layout

Step 4: IDS Layout

The selected radicals are arranged using an IDS (Ideographic Description Sequence) layout template. Twelve standard compositions are available:

• ▪Two-component: ⿰ (left-right), ⿱ (top-bottom)
• ▪Three-component: ⿲ (left-middle-right), ⿳ (top-middle-bottom)
• ▪Enclosures: ⿴ ⿵ ⿶ ⿷ ⿸ ⿹ ⿺ (various open/closed surrounds)
• ▪Overlay: ⿻ (components overlapping)

Layout selection is influenced by the character type — compound semantics prefer side-by-side, while enclosures are used for entity-type characters.

Step 5: Stroke Assembly

Each radical's KAGE stroke data is remapped into its assigned layout slot. Stroke coordinates are scaled and translated from the radical's native 200×200 grid to the slot's bounding box within the final 200×200 glyph space.

radical.strokeData → remapStrokesToSlot(slot) → assembled stroke string

Individual radical stroke strings are concatenated with $ delimiters to form the complete glyph data.

Step 6: Rendering

The assembled stroke data string is passed to the KAGE engine:

1. ▪Parse: Split by $, each segment becomes a StrokeData tuple
2. ▪Render: Mincho (default) or Gothic font renderer processes each stroke
3. ▪Output: Polygon outlines → SVG path data → inline <svg> element

▸Dual Readings

Every generated edenji also receives two readings:

• ▪Onyomi: Monosyllabic CVC form, selected from the seed's phonological neighborhood
• ▪Kunyomi: Polysyllabic 2–3 syllable form, independently generated

Reading selection encodes syntactic information — onyomi for compounds, kunyomi for standalone use.

▸Determinism

The entire pipeline is deterministic: seed → hash → PRNG → radicals → layout → strokes → SVG. No randomness enters the system. This means:

• ▪The same word always produces the same glyph
• ▪Glyphs can be cached by seed string
• ▪The glossary stores seed strings rather than glyph data