Design Generative Art

Algorithmic Art Skill

Degree of freedom: MIXED. Algorithm and palette [HIGH freedom]; existing pipeline inventory and seed reproducibility [LOW freedom — run exactly].

Create generative, procedural, and mathematical art using code. Transform algorithms into visual experiences.

How to reason

1. Inventory — existing p5 / canvas / noise utilities
2. Seed — deterministic RNG; same seed = same pixels
3. Parameterize — density, palette, scale, complexity
4. Validate — resolution-independent; reduced-motion on loops

Worked example

Inventory: no p5; one getContext('2d') chart canvas — do not hijack it. Seed: mulberry32(42) drives a new src/components/art/ flow field. Parameterize: density, palette (ocean), scale; slider re-renders from the same seed. Validate: seed 42 twice matches; prefers-reduced-motion stops the RAF loop.

Self-critique before reporting

• No pipeline clash — did not overwrite an existing canvas/WebGL path
• Reproducible — same seed paints identical output
• Controls matter — each param changes the picture, not just a label
• Right owner — data charts → data-visualization; UI illustration → design-frontend; WebGL/3D → enhance-web-web3d

Check existing first [LOW freedom — run exactly]

Before creating ANY generative art, verify:

1. Check for existing creative coding setup:

cat package.json | grep -i "p5\|three\|canvas\|pixi\|paper"
ls -la src/components/art/ src/components/generative/ 2>/dev/null

2. Check for existing canvas/WebGL usage:

rg "Canvas|useFrame|getContext.*2d|WebGL" --type tsx -l

3. Check for existing noise/random utilities:

rg "simplex\|perlin\|noise\|seedrandom" --type ts

Why: Don't conflict with existing rendering pipelines or duplicate utility code.

Core Principles [HIGH freedom]

1. Seeded Randomness

Every piece should be reproducible with a seed:

// Deterministic random number generator
function mulberry32(seed: number) {
 return function() {
 let t = seed += 0x6D2B79F5
 t = Math.imul(t ^ t >>> 15, t | 1)
 t ^= t + Math.imul(t ^ t >>> 7, t | 61)
 return ((t ^ t >>> 14) >>> 0) / 4294967296
 }
}

// Usage
const rng = mulberry32(42) // Same seed = same output
const value = rng() // 0-1 deterministic random

2. Parameterized Generation

Make art controllable via parameters:

interface ArtParams {
 seed: number
 density: number // 0-1
 palette: string[]
 scale: number
 speed: number
 complexity: number // 0-1
}

3. Resolution Independence

Design for any canvas size:

// Normalize coordinates to 0-1 range
const nx = x / width
const ny = y / height
// Then scale to canvas
const px = nx * canvas.width
const py = ny * canvas.height

Techniques

Flow Fields

function createFlowField(cols: number, rows: number, seed: number) {
 const rng = mulberry32(seed)
 const field: number[][] = []

 for (let y = 0; y < rows; y++) {
 field[y] = []
 for (let x = 0; x < cols; x++) {
 // Perlin-like noise using layered sine waves
 const angle = Math.sin(x * 0.1) * Math.cos(y * 0.1) * Math.PI * 2
 + rng() * 0.5
 field[y][x] = angle
 }
 }
 return field
}

function drawFlowField(ctx: CanvasRenderingContext2D, field: number[][], params: ArtParams) {
 const cellW = ctx.canvas.width / field[0].length
 const cellH = ctx.canvas.height / field.length

 // Spawn particles and follow flow
 for (let i = 0; i < params.density * 1000; i++) {
 let x = rng() * ctx.canvas.width
 let y = rng() * ctx.canvas.height

 ctx.beginPath()
 ctx.moveTo(x, y)
 ctx.strokeStyle = params.palette[Math.floor(rng() * params.palette.length)]
 ctx.globalAlpha = 0.3

 for (let step = 0; step < 100; step++) {
 const col = Math.floor(x / cellW)
 const row = Math.floor(y / cellH)

 if (col < 0 || col >= field[0].length || row < 0 || row >= field.length) break

 const angle = field[row][col]
 x += Math.cos(angle) * params.scale
 y += Math.sin(angle) * params.scale
 ctx.lineTo(x, y)
 }

 ctx.stroke()
 }
}

Recursive Subdivision

function subdivide(
 ctx: CanvasRenderingContext2D,
 x: number, y: number, w: number, h: number,
 depth: number, maxDepth: number, rng: () => number,
 palette: string[]
) {
 if (depth >= maxDepth || rng() < 0.15) {
 // Draw leaf
 ctx.fillStyle = palette[Math.floor(rng() * palette.length)]
 ctx.globalAlpha = 0.6 + rng() * 0.4
 ctx.fillRect(x + 1, y + 1, w - 2, h - 2)
 return
 }

 // Random split direction and position
 const horizontal = rng() > 0.5
 const split = 0.3 + rng() * 0.4 // 30-70% split

 if (horizontal) {
 const splitY = y + h * split
 subdivide(ctx, x, y, w, splitY - y, depth + 1, maxDepth, rng, palette)
 subdivide(ctx, x, splitY, w, y + h - splitY, depth + 1, maxDepth, rng, palette)
 } else {
 const splitX = x + w * split
 subdivide(ctx, x, y, splitX - x, h, depth + 1, maxDepth, rng, palette)
 subdivide(ctx, splitX, y, x + w - splitX, h, depth + 1, maxDepth, rng, palette)
 }
}

Circle Packing

interface Circle {
 x: number; y: number; r: number; color: string
}

function circlePacking(
 width: number, height: number,
 maxCircles: number, maxRadius: number,
 rng: () => number, palette: string[]
): Circle[] {
 const circles: Circle[] = []
 let attempts = 0
 const maxAttempts = maxCircles * 50

 while (circles.length < maxCircles && attempts < maxAttempts) {
 attempts++
 const candidate = {
 x: rng() * width,
 y: rng() * height,
 r: 2,
 color: palette[Math.floor(rng() * palette.length)]
 }

 // Grow until collision
 let valid = true
 while (valid && candidate.r < maxRadius) {
 candidate.r += 1
 for (const other of circles) {
 const dist = Math.hypot(candidate.x - other.x, candidate.y - other.y)
 if (dist < candidate.r + other.r + 2) {
 candidate.r -= 1
 valid = false
 break
 }
 }
 // Check bounds
 if (candidate.x - candidate.r < 0 || candidate.x + candidate.r > width ||
 candidate.y - candidate.r < 0 || candidate.y + candidate.r > height) {
 candidate.r -= 1
 valid = false
 }
 }

 if (candidate.r > 2) circles.push(candidate)
 }
 return circles
}

L-Systems (Fractal Trees/Plants)

interface LSystem {
 axiom: string
 rules: Record<string, string>
 angle: number
 length: number
 iterations: number
}

const fractalTree: LSystem = {
 axiom: 'F',
 rules: { 'F': 'FF+[+F-F-F]-[-F+F+F]' },
 angle: 25,
 length: 4,
 iterations: 4,
}

function generateLSystem(system: LSystem): string {
 let current = system.axiom
 for (let i = 0; i < system.iterations; i++) {
 current = current.split('').map(c => system.rules[c] || c).join('')
 }
 return current
}

function drawLSystem(ctx: CanvasRenderingContext2D, system: LSystem, startX: number, startY: number) {
 const instructions = generateLSystem(system)
 const stack: { x: number; y: number; angle: number }[] = []
 let x = startX, y = startY, angle = -90 // Start pointing up

 ctx.beginPath()
 ctx.moveTo(x, y)

 for (const char of instructions) {
 switch (char) {
 case 'F':
 const nx = x + Math.cos(angle * Math.PI / 180) * system.length
 const ny = y + Math.sin(angle * Math.PI / 180) * system.length
 ctx.lineTo(nx, ny)
 x = nx; y = ny
 break
 case '+': angle += system.angle; break
 case '-': angle -= system.angle; break
 case '[': stack.push({ x, y, angle }); break
 case ']':
 const state = stack.pop()!
 x = state.x; y = state.y; angle = state.angle
 ctx.moveTo(x, y)
 break
 }
 }
 ctx.stroke()
}

React Component Pattern

'use client'
import { useEffect, useRef, useState, useCallback } from 'react'

interface GenerativeArtProps {
 seed?: number
 width?: number
 height?: number
 palette?: string[]
 className?: string
}

export function GenerativeArt({
 seed = Date.now(),
 width = 800,
 height = 600,
 palette = ['#264653', '#2a9d8f', '#e9c46a', '#f4a261', '#e76f51'],
 className,
}: GenerativeArtProps) {
 const canvasRef = useRef<HTMLCanvasElement>(null)
 const [currentSeed, setCurrentSeed] = useState(seed)

 const render = useCallback(() => {
 const canvas = canvasRef.current
 if (!canvas) return
 const ctx = canvas.getContext('2d')!
 const rng = mulberry32(currentSeed)

 // Clear
 ctx.fillStyle = '#1a1a2e'
 ctx.fillRect(0, 0, width, height)

 // Your generative algorithm here
 drawFlowField(ctx, createFlowField(40, 30, currentSeed), {
 seed: currentSeed,
 density: 0.8,
 palette,
 scale: 2,
 speed: 1,
 complexity: 0.7,
 })
 }, [currentSeed, width, height, palette])

 useEffect(() => { render() }, [render])

 return (
 <div className={className}>
 <canvas
 ref={canvasRef}
 width={width}
 height={height}
 className="rounded-lg"
 style={{ maxWidth: '100%', height: 'auto' }}
 />
 <div className="flex gap-2 mt-4">
 <button
 => setCurrentSeed(Date.now())}
 className="px-4 py-2 bg-primary text-primary-foreground rounded-lg"
 >
 Regenerate
 </button>
 <input
 type="number"
 value={currentSeed}
 => setCurrentSeed(Number(e.target.value))}
 className="px-3 py-2 border rounded-lg w-32"
 aria-label="Seed value"
 />
 </div>
 </div>
 )
}

Color Palettes

// Curated palettes for generative art
const PALETTES = {
 // Warm
 sunset: ['#ff6b6b', '#feca57', '#ff9ff3', '#54a0ff', '#5f27cd'],
 autumn: ['#d35400', '#e67e22', '#f39c12', '#2c3e50', '#ecf0f1'],

 // Cool
 ocean: ['#0c2461', '#1e3799', '#4a69bd', '#6a89cc', '#82ccdd'],
 forest: ['#1b4332', '#2d6a4f', '#40916c', '#52b788', '#74c69d'],

 // Monochrome
 ink: ['#000000', '#1a1a1a', '#333333', '#4d4d4d', '#666666'],
 paper: ['#f5f0e8', '#ede4d4', '#e5d9c0', '#ddc9a3', '#d4ba87'],

 // Vibrant
 neon: ['#ff00ff', '#00ffff', '#ff0066', '#66ff00', '#ffff00'],
 candy: ['#ff6f91', '#ff9671', '#ffc75f', '#f9f871', '#d4fc79'],

 // Japanese-inspired
 wabi: ['#2c1810', '#5c3a2e', '#b5651d', '#daa06d', '#f5deb3'],
 sakura: ['#ffb7c5', '#ff69b4', '#c71585', '#8b008b', '#4a0028'],
}

Animation Loop

function animatedArt(canvas: HTMLCanvasElement, params: ArtParams) {
 const ctx = canvas.getContext('2d')!
 let frame = 0
 let animationId: number

 function loop() {
 frame++
 const t = frame * params.speed * 0.01

 // Semi-transparent overlay for trails
 ctx.fillStyle = 'rgba(0, 0, 0, 0.02)'
 ctx.fillRect(0, 0, canvas.width, canvas.height)

 // Animated elements
 for (let i = 0; i < 50; i++) {
 const x = canvas.width / 2 + Math.cos(t + i * 0.5) * 200
 const y = canvas.height / 2 + Math.sin(t * 0.7 + i * 0.3) * 200
 const r = 2 + Math.sin(t + i) * 1

 ctx.beginPath()
 ctx.arc(x, y, r, 0, Math.PI * 2)
 ctx.fillStyle = params.palette[i % params.palette.length]
 ctx.globalAlpha = 0.8
 ctx.fill()
 }

 animationId = requestAnimationFrame(loop)
 }

 loop()
 return () => cancelAnimationFrame(animationId)
}

Export & Sharing

// Export canvas as PNG
function exportPNG(canvas: HTMLCanvasElement, filename: string) {
 const link = document.createElement('a')
 link.download = `${filename}-${Date.now()}.png`
 link.href = canvas.toDataURL('image/png')
 link.click()
}

// Export as SVG (for vector output)
function exportSVG(svgElement: SVGSVGElement, filename: string) {
 const serializer = new XMLSerializer()
 const svgString = serializer.serializeToString(svgElement)
 const blob = new Blob([svgString], { type: 'image/svg+xml' })
 const link = document.createElement('a')
 link.download = `${filename}-${Date.now()}.svg`
 link.href = URL.createObjectURL(blob)
 link.click()
}

Related Skills

• enhance-web-web3d — WebGL, Three.js, shaders for 3D generative art
• design-motion — Animation patterns for interactive pieces
• design-canvas — Print-quality visual design philosophy
• data-visualization — Data-driven generative compositions

Validation [LOW freedom — do not skip]

After creating algorithmic art:

1. Reproducibility → Same seed produces identical output
2. Performance → 60fps for animated pieces
3. Resolution → Looks good at target export size
4. Palette → Colors work together harmoniously
5. Parameters → Controls produce meaningful visual changes
6. Export → PNG/SVG export works correctly
7. Accessibility → Animated art respects prefers-reduced-motion