About

Procedural landscape generation uses layered noise functions to synthesize terrain profiles that mimic geological erosion patterns. Each pixel row on the canvas represents a depth layer rendered with aerial perspective - distant ranges lose saturation by roughly 15% per layer while shifting toward the horizon color. The generator combines fractal Brownian motion (summing n octaves of coherent noise at halving amplitudes) with midpoint displacement to produce ridge lines that avoid the artificial periodicity of simple sine waves. Results approximate real orographic silhouettes but assume uniform erosion. They will not replicate specific geological formations like karst towers or fjords.

Seed-based determinism means every landscape is reproducible. Record the seed value and you can regenerate identical output on any device. The atmospheric model uses a simplified Rayleigh scattering approximation - warm tones near the horizon at low sun angles, cooler zenith overhead. Vegetation density follows a slope threshold: trees avoid gradients steeper than roughly 0.6. Water bodies are placed below a configurable sea-level parameter with basic specular reflection. This tool does not simulate weather systems or volumetric clouds.

Formulas

Terrain height at horizontal position x is computed via fractal Brownian motion - a weighted sum of noise octaves:

H(x) = n−1∑i=0 A ⋅ pⁱ ⋅ noise(x ⋅ f ⋅ Lⁱ)

where A is base amplitude, p is persistence (typically 0.5), f is base frequency, L is lacunarity (typically 2.0), and n is octave count. The noise function returns coherent pseudo-random values in [−1, 1] using a permutation table seeded by the user-provided seed S.

Aerial perspective blends each layer’s color C_layer toward the horizon color C_horizon:

C_final = lerp(C_layer, C_horizon, dd_max)

where d is the layer depth index and d_max is total layer count. Sun angle θ drives the sky gradient color stops: at θ < 15° warm sunset tones dominate; at θ > 45° the zenith shifts to deep blue.

Reference Data

Biome Preset	Terrain Layers	Avg. Roughness	Vegetation	Water Level	Sky Type	Typical Palette	Real-World Analog
Alpine	6	0.65	Sparse conifers	20%	Clear / Haze	Blue-grey, white peaks	Swiss Alps, Rockies
Desert	4	0.30	None / Cacti	0%	Harsh sun	Amber, ochre, sand	Sahara, Mojave
Tropical	5	0.50	Dense palms	35%	Humid haze	Emerald, turquoise	Southeast Asia, Caribbean
Arctic	5	0.55	None	30% (ice)	Overcast	White, pale blue, grey	Antarctica, Svalbard
Rolling Hills	4	0.25	Grass, scattered trees	10%	Soft gradient	Green, golden	Tuscany, English countryside
Canyon	5	0.80	Minimal shrubs	5%	Clear sky	Rust, terracotta, cream	Grand Canyon, Wadi Rum
Forest	5	0.40	Dense mixed	15%	Filtered light	Deep green, brown, mist	Black Forest, Taiga
Coastal	4	0.35	Moderate	45%	Maritime haze	Teal, sandy, sky blue	Amalfi, Big Sur
Volcanic	5	0.75	Sparse	10%	Smoky	Dark grey, orange glow	Iceland, Hawaii
Savanna	3	0.20	Acacia-style scattered	5%	Warm gradient	Gold, sienna, dusty green	Serengeti, Outback
Fjord	6	0.70	Sparse conifers	50%	Overcast	Steel blue, dark green	Norway, New Zealand
Plateau	4	0.15	Low scrub	0%	Clear	Tan, muted purple	Tibet, Altiplano

Frequently Asked Questions

The seed initializes a deterministic pseudo-random number generator (a linear congruential generator with modulus 2³¹ − 1). Every call to the noise function pulls from this sequence. Given identical seed, octave count, persistence, and canvas dimensions, the output is bit-identical. If you change canvas size, the landscape shape is preserved but pixel resolution differs.

This simulates aerial (atmospheric) perspective. Real atmosphere scatters short-wavelength light (Rayleigh scattering), causing distant objects to shift toward blue-grey and lose contrast. The generator applies a linear interpolation from full-saturation foreground color to the horizon tone, with blending weight proportional to layer depth index divided by total layers. Increasing layer count amplifies the effect.

Two parameters interact: persistence and octave count. Persistence (variable p, range 0.1-0.9) controls how much amplitude each successive octave retains. Low persistence (≈ 0.3) produces smooth, gentle terrain. High persistence (≈ 0.7) preserves high-frequency detail, creating jagged peaks. Octave count adds layers of detail - 2 octaves yield broad shapes, 6+ octaves add fine craggy texture. The biome presets configure both values to match geological archetypes.

Yes. The output is procedurally generated from mathematical functions with no copyrighted source material. The PNG export contains no metadata linking to this tool. You own the pixel data. However, note that simple procedural landscapes lack the complexity of hand-painted art - they work well for backgrounds, placeholders, album covers, and game assets but may look repetitive at large scales without post-processing.

Trees are placed using a Poisson-disk-like scatter: candidate positions are generated along the terrain surface, then rejected if the local terrain slope exceeds a threshold (default 0.6 gradient). A minimum spacing parameter prevents overlap. Vertical size varies ±20% randomly. This avoids grid artifacts and uniform rows. Density is controlled per-biome - desert has near-zero density, forest has maximum packing within the spacing constraint.

The canvas renders at the displayed pixel dimensions, which adapt responsively to the container width (typically 1200×600 on desktop, scaling down on mobile). For high-resolution export, the generator can render at 2× device pixel ratio. Aspect ratio is fixed at 2:1 (panoramic), matching common wallpaper and banner formats. The exported PNG preserves exact canvas pixel dimensions.