Photo Puzzle - fotojockovic.com/puzzle
======================================
Turns a photo into a puzzle (30 types), lets you play it in the browser, send it to a friend, and download exact
cut files (SVG, DXF, PDF) and a print file (PDF with bleed, crop marks and CutContour lines). Everything runs in the
browser; no photo leaves the device.

Our own work, written for this tool (source: src-rs/, MIT licence), with no third-party code or crates:
  - pz.wasm: the geometry engine, Rust compiled to WebAssembly and run in worker.js:
      * jigsaw knob ("tab") curves in 8 styles from Hermite nodes -> cubic Beziers, with size / neck / variety /
        waviness controls; knobs that would collide with a neighbour, the outline or a whimsy are detected and
        replaced, and self-crossing curves are rejected
      * 30 puzzle generators (see PUZZLE-TYPES.txt): square, brick, diamond, hexagon, triangle, rhombille and
        Penrose (Robinson-triangle deflation) lattices, Voronoi cells with Lloyd relaxation, polar (round) grid,
        Archimedean spiral, shaped outlines, letters, whimsies, games (sliding, rotating, swap, strips, rings,
        tangram, pentomino with a randomised exact-cover search)
      * planar arrangement: all curves flattened for topology only, crossings found with a uniform grid, curves
        split at exact Bezier parameters and welded to shared vertices, faces traced on a half-edge structure,
        bridges for holes, merging of slivers; every cut edge keeps its exact lines/cubics for export
      * self-check of every puzzle (and a re-try with the next seed if it fails): pieces' areas add up to the
        outline area, no crossings, no loose ends, exact size
      * signed distance field atlas of all pieces for the WebGL renderer and for picking
      * cut files: SVG (inner cuts red, outline blue, frame green, numbers as an engrave layer), DXF R12
        (layers CUT_INNER, CUT_OUTLINE, CUT_FRAME, ENGRAVE), PDF; shared edges cut once, chained into long paths,
        inner cuts before the outline; kerf compensation, separate-pieces layout, frame and base board
      * print PDF: the photo with bleed (edge-extended where needed), crop marks, cut lines as the CutContour
        Separation spot colour (or thin black), page sizes, a mirrored back side with piece numbers or a second
        photo for duplex printing
      * subject outline for silhouette puzzles: OKLab k-means of the border colours, Otsu threshold, morphology,
        Moore boundary tracing, smoothing (no machine learning)
      * letter outlines: a small TrueType reader (cmap 4, glyf, composites) used at build time to pack the glyphs
        we need into font/name.bin
  - gl.js: WebGL2 renderer (instanced pieces, bevel lighting and soft shadows from the distance field).
  - app.js, worker.js, i18n.js, lang/*.js, index.html: the page, UI and texts.

Published methods the engine builds on (ideas only; no code was copied):
  - R. Penrose's P3 tiling via Robinson triangle subdivision (as popularised by J. Preshing's description).
  - S. Lloyd, "Least squares quantization in PCM" (Lloyd relaxation of Voronoi cells).
  - B. Ottosson, "A perceptual color space for image processing" (OKLab), 2020 - public domain formulas.
  - N. Otsu, "A threshold selection method from gray-level histograms", 1979.
  - S. Wang's bound for the subdivision of Bezier curves; de Casteljau splitting; Hermite-to-Bezier conversion.
  - The TrueType format as documented by Apple and Microsoft (OpenType specification).

Sample pictures (img/*.jpg, resized). The Metropolitan Museum of Art, Open Access, CC0 1.0
(isPublicDomain: true in the Met Collection API, checked 2026-10-16):
  img/wave.jpg  - Katsushika Hokusai, Under the Wave off Kanagawa (The Great Wave), ca. 1830-32, object 45434
  img/harv.jpg  - Pieter Bruegel the Elder, The Harvesters, 1565, object 435809
  img/wheat.jpg - Vincent van Gogh, Wheat Field with Cypresses, 1889, object 436535
  img/jar.jpg   - Jacob Scholl, Covered jar, ca. 1830-40, object 2412

Font for name puzzles (font/name.bin: outlines of Latin, Greek and Cyrillic letters and digits extracted from the
font, no font file is shipped as such): Fira Sans Black, Copyright (c) 2012-2015 The Mozilla Foundation and
Telefonica S.A., SIL Open Font License 1.1 (full text: font/OFL.txt). The extracted outlines are a Modified
Version under the OFL and keep its licence; they are not sold on their own.

UI fonts, loaded from Google Fonts, SIL Open Font License 1.1: Inter Tight (Rasmus Andersson et al.), Instrument
Serif (Instrument).

Shared helpers /static/tools/common/i18n.js, rail.js and sheets.js: our own. Translations of the interface: our own.

Used only to build and test, never shipped to the browser: Rust and cargo (MIT/Apache-2.0); binaryen (Apache-2.0)
for wasm-opt; puppeteer-core (Apache-2.0) and Google Chrome for screenshots and tests; poppler pdftoppm (GPL, run
locally to look at the PDFs only).

No GPL code is shipped. No machine-learning model is used.

Self-hosted web fonts (2026-10-19): Inter Tight, Instrument Serif. SIL Open Font License 1.1. The woff2 files are the ones Google Fonts
serves (latin, latin-ext, cyrillic, greek subsets where the family has them), stored on our own server under
/static/tools/common/fonts/<family>/ with each family's licence text (OFL.txt) next to them; stylesheet
/static/tools/common/fonts/ui.css. No font request goes to Google.