Noise

This page will explain how the Perlin Noise Generator up to at least Beta 1.7.3 worked.

  1. Perlin Noise
  2. Simplex Noise
  3. Modifications made for Minecraft
  4. Noise Octaves
  5. Further reading

Large parts of this are either missing or super simplified. Minecraft sometimes runs a function a certain way depending on how many inputs are provided (2D vs 3D).

Perlin Noise

The algorithm used by most of Minecraft’s Terrain is based on “Improved Perlin Noise”, first published in 2002, which is a slightly tweaked version of the original Perlin Noise Algorithm from 1985, both by Ken Perlin.

The official paper can be found here, however a more accessible option is the Wikipedia page on Perlin noise.

The Farlands occur here due to a signed 64-Bit Floating-point number being converted to a signed 32-Bit Integer, resulting in the same permutation from the permutation table being chosen repeatedly.

For an example that you can reference for your purposes, check out the BetrockServer Source Code.

Simplex Noise

Minecraft utilizes a slightly modified implementation of Simplex Noise for its Biome values (Temperature & Humidity). As with Perlin noise, more can be found about this algorithm on Wikipedia. It was also developed by Ken Perlin.

For an example that you can reference for your purposes, check out the BetrockServer Source Code.

Modifications made for Minecraft

The main differences that arise in the Perlin and Simplex Noise generators comes from some slight tweaks that were made when implementing them.

  1. An inital pseudorandom offset is assigned to each of the coordinates of the noise generator coordinate = nextDouble() * 256.0;
  2. For shuffling the permutation table, once again, Javas PRNG is used

Noise Octaves

A common practice when using any noise is to use multiple octaves/layers of it to improve fine detail, resulting in a more realistic or natural appearance.

It boils down to halving the size of the noise, while also halving the amount of noise each additional octave adds.

X, Y and Z values are passed as input. Then appropriately scaled noise is sampled, with the result scaled by the multiplier.

double total = 0.0;
double mult  = 1.0;
for (int i = 0; i < octaves; ++i) {
    total += perlinNoise(x / mult, y / mult, z / mult) * mult;
    mult  *= 2.0;
}
return total;

Some of the octave generation functions work a little differently than others. Check out the BetrockServer Source Code for more info.

Further reading

From here on out, one can read about how the rest of the world generation is done. Check out the page for the World Generator.