Each piece is a small swatch of cloth. Warp ribbons run down the canvas and weft ribbons run across it. Where they cross, a weave structure decides which one goes on top: plain weave, a twill (sometimes broken into a herringbone), a basket, a satin with long floats, or a draft written the way a hand-loom is set up, with a threading, a treadling and a tie-up. Every ribbon rises over the crossings it covers and dips under the others, so light and cast shadows show the depth. Coloured stripes in the warp and the weft then meet through the structure as checks and plaids. The cloth may lie flat, bend gently or ripple, and its ends may fray. Press New seed for a fresh hash, then share the URL to show someone that exact piece.
Traits
The weights were set before any drawing code was written. They were then checked against a census of 2,000 seeds.
| Trait | Values (weight) |
|---|---|
| Palette | Linen 22 · Madder 20 · Fjord 16 · Moss 14 · Saffron 14 · Nocturne 14 |
| Pattern | Plain 22 · Twill 26 · Basket 16 · Satin 14 · Draft 22 |
| Scale | Fine 25 · Medium 45 · Coarse 30 |
| Warp | Straight 40 · Bent 40 · Rippled 20 |
| Stripes | Solid 25 · Banded 35 · Plaid 40 |
Credits
- Structures: plain weave, twills, basket weave and satins, together with the loom draft notation (threading, treadling and tie-up) that hand-weavers have used for centuries and that Jacquard’s punched-card loom later automated. The drafts here are generated from the hash, and floats longer than five crossings are broken up.
- The stripe sequences are mirrored about a pivot the way tartan setts are, but they are generated. None of them copies a real tartan, and the palettes are original.
- Rendering: every pixel is mapped back into flat cloth space through a low-frequency displacement field, then shaded from a smoothstep height profile along each ribbon, with analytic edge coverage for anti-aliasing.
- Randomness: sfc32 by Chris Doty-Humphrey. The hash seeding, named sub-streams and verification follow Camille Roux’s genart skill (MIT).
- The code was written with an AI assistant. It draws everything itself: no AI-generated images.