Morphogenesis Lab · live Gray–Scott culture

MORPHÊ

Two chemicals, one rule, and no plan. Drag a seed onto the plate and the field decides its own form — spots, stripes, coral, division — pattern condensing out of nothing but diffusion.

Drag inside the plate to inoculate. It grows on its own between touches.

Plate 01 · live culture
Feed f0.0545
Kill k0.0620
Step Δt1.00
Iteration0
01 · The plate

A culture that grows whether or not you touch it

MORPHÊ keeps one plate warm at all times. There is no image loaded, no animation looping — every frame is computed from the last by the same two-chemical rule, on the GPU, right now.

Alan Turing asked a strange question in 1952: how does a uniform ball of cells — identical everywhere — decide where to put a stripe? His answer was that pattern needs no blueprint. Two substances, one that spreads slowly and reinforces itself, one that spreads fast and erases, are enough. Left alone, they refuse to stay flat.

The plate above runs that idea as a real simulation. An activator and an inhibitor diffuse across a 512-cell grid; where the activator wins it darkens to culture teal, where the two fight it flares coral. Nudge it with a seed and you tip the balance — the rest is chemistry finding its own equilibrium.

Nothing here is drawn. Wipe your cursor across the agar and you are not painting a picture; you are raising the local concentration of one reactant and letting diffusion carry the consequence outward for the next thousand steps.

262,144Cells simulated per plate
~8Reaction steps computed each frame
2Reactants — one activator, one inhibitor
1952Turing's morphogenesis paper
02 · Four regimes

Change two numbers, change the whole animal

Feed rate f and kill rate k are the only dials. A few thousandths apart and the same culture grows a leopard, a fingerprint, a reef, or a colony that splits forever. Load one and watch the plate re-decide.

Leopard spots

Isolated dots that space themselves evenly and hold. The classic stable Turing spot — self-limiting, tidy, patient.

f0.035k0.062

Fingerprint stripes

Spots stretch and fuse into wandering ridges — the labyrinth of a fingertip, drawn by nothing but competing diffusion speeds.

f0.026k0.051

Coral labyrinth

Branching fronts that keep budding new tips as they crawl — a reef building itself, never quite filling the plate.

f0.055k0.062

Mitosis

Every spot grows until it pinches in two, and the daughters do the same. A colony that divides without end — life's oldest trick, in two lines of maths.

f0.037k0.065
03 · The method

The whole organism is these two lines

The Gray–Scott model tracks concentrations of an activator U and inhibitor V. Each cell reads its neighbours, applies the reaction, and writes the result — the GPU does all 262,144 of them at once, then reads the new plate back in as the input for the next step.

Diffuse

Each reactant bleeds into its neighbours by a Laplacian stencil. The activator spreads slowly, the inhibitor twice as fast — the asymmetry is the entire source of pattern.

React

Where activator and inhibitor meet, the reaction U·V² converts one to the other. Feed replenishes U everywhere; kill removes V. Balance decides the form.

Ping-pong

Two GPU textures take turns: read from A, write to B, then swap. No frame is stored twice; the pattern is only ever the running total of every step since you seeded it.

# activator — slow diffusion, consumed by the reaction, topped up by feed
∂U/∂t = Dᵤ∇²U UV² + f(1 U)
# inhibitor — fast diffusion, produced by the reaction, removed by kill
∂V/∂t = Dᵥ∇²V + UV² (f + k)V
04 · Residency

Six weeks. One plate. A pattern nobody has grown before.

MORPHÊ runs a short residency for artists, biologists and shader people who want to work with self-organising systems for real — no metaphors, no stock footage. You get a workstation, a parameter space, and the patience to sit with a plate until it surprises you. Applications open twice a year.

Request the residency brief

Perth · next intake opens 1 September · bring a hypothesis