Delaunay Triangulation Generator

Continue in the tool

More pages for this tool

triangulation for generative poster

Delaunay triangulation for a generative poster

Generative poster art needs a quality pure randomness cannot give: the ability to return to the exact same composition after testing variations. A fixed text seed pins the pseudorandom sequence behind a Delaunay triangulation, turning the mesh into a versionable starting point instead of a lost dice roll.

Seed, density and the triangle ceiling per poster

The text seed ("poster-01", for example) initializes a deterministic pseudorandom generator that decides every point position before the triangulation runs. Changing a single character in the seed changes the entire sequence of generated numbers, and therefore every point and triangle that follows: there is no middle ground between "same seed, same mesh" and "different seed, completely distinct mesh".
With seed "poster-01" and 72 points on the default 720 by 420 canvas, the verified mesh has 132 triangles and about 21 KB of SVG; the theoretical 2N-2 ceiling for 72 points is 142 triangles, and the missing 10 correspond to the composition's convex hull points, the outer edge the triangulation does not close from inside.
Point density and jitter (random position offset) are the two parameters that change the piece's character without switching algorithms: more points raise the triangle count and perceived detail, more jitter pulls the mesh away from a regular grid and toward an organic composition. Both can be versioned alongside the seed to reproduce the same visual direction in later iterations of the same poster.

Three versions of the same poster direction

Poster base version

Input: seed: poster-01, points: 72, width: 720, height: 420, jitter: 30
Expected output: 132 triângulos, SVG de aproximadamente 21 KB

This is the reference mesh: any designer with the same seed and parameters recreates this exact result.

Variant with reduced jitter

Input: seed: poster-01, points: 72, width: 720, height: 420, jitter: 5
Expected output: Mesma seed, malha mais regular e menos orgânica que a versão base

Lowering jitter pulls the points toward a grid, useful for a more geometric variant of the same poster without changing the seed.

New direction for the same series

Input: seed: poster-02, points: 72, width: 720, height: 420, jitter: 30
Expected output: Malha totalmente distinta da poster-01, mesmo com parâmetros idênticos

Changing only the seed, while keeping density and jitter, is how to generate a poster family with the same "style" but unique compositions.

Full tool FAQ

What makes Delaunay different from connecting points randomly?

The empty-circumcircle rule: no point lies inside the circle through a triangle's vertices. That maximizes the smallest angle and avoids extremely narrow triangles, producing a more balanced mesh.

Does the seed guarantee the same result?

Are more points always better?

Can Delaunay be used in UI?

Can the SVG be edited later?

Is the triangulation scientific-grade?

Can the result become heavy?

Does generation use a backend?

Frequently asked questions

Why is the seed better than saving the final SVG as a reference?

Because the seed stores the recipe, not the result: with it you can reopen the composition and change just one parameter, like point density or poster dimensions, while keeping the same placement logic. Saving only the final SVG locks the piece as a static file, with no way back to adjust a single parameter.

Does raising points from 72 to 120 change the whole composition or just add detail?

It changes the whole composition: the pseudorandom sequence the seed produces depends on the requested point count, so every new point total redraws every point's position, not just appends new ones at the end. To add detail while preserving the overall layout, raise the count gradually and compare visually.

Can I predict how many triangles a seed will produce before running it?

Only the ceiling: 2N-2 for N points, without the convex hull discount. The exact triangle count depends on how many points land on the composition's outer boundary (the convex hull), which varies with seed and jitter, so the real value always sits a bit below the theoretical ceiling.

Does the same seed produce the same result if I change the poster width or height?

No: each point position is calculated relative to the canvas dimensions, so changing width or height repositions every point proportionally to the new area, even with the seed's pseudorandom sequence untouched. To reopen the exact same mesh, keep seed, point count, dimensions and jitter identical.

Delaunay Triangulation Generator

Continue in the tool

More pages for this tool

Delaunay triangulation for a generative poster

Seed, density and the triangle ceiling per poster

Three versions of the same poster direction

Poster base version

Variant with reduced jitter

New direction for the same series

Full tool FAQ

Frequently asked questions

Why is the seed better than saving the final SVG as a reference?

Does raising points from 72 to 120 change the whole composition or just add detail?

Can I predict how many triangles a seed will produce before running it?

Does the same seed produce the same result if I change the poster width or height?

Next steps

Delaunay Triangulation Generator

Continue in the tool

More pages for this tool

Delaunay triangulation for a generative poster

Seed, density and the triangle ceiling per poster

Three versions of the same poster direction

Poster base version

Variant with reduced jitter

New direction for the same series

Full tool FAQ

Frequently asked questions

Why is the seed better than saving the final SVG as a reference?

Does raising points from 72 to 120 change the whole composition or just add detail?

Can I predict how many triangles a seed will produce before running it?

Does the same seed produce the same result if I change the poster width or height?

Next steps