vnoise is a pure-Python, Numpy-based, vectorized port of the noise library. It currently implements the Perlin noise functions in 1D, 2D and 3D version (noise also implements simplex noise).

vnoise was started because the original noise library is no longer supported and binaries for recent versions of Python are unavailable, making it hard to install for non-technical users. vnoise does not suffer from the same issue since it is written in pure Python.

For scalar input (e.g. when computing noise values one at a time), yes (~300-2000x slower depending on the conditions).

vnoise deals with this by offering a vectorized API to compute large numbers of noise values in a single call. Since is uses Numpy internally, its performance can match or exceed the original library (0.3-4x faster depending on the conditions).

$ pip install vnoise

Basic example:

>>> import vnoise
>>> noise = vnoise.Noise()
>>> noise.noise1(0.5)
0.0
>>> noise.noise1(0.1)
0.09144000000000001
>>> noise.noise2(0.1, 0.3)
0.09046282464000001
>>> noise.noise3(0.1, 0.3, 0.7)
0.27788822071249925

The noiseX() functions also accept sequences as arguments:

>>> import numpy as np
>>> noise.noise2([0.1, 0.2, 0.3], np.linspace(0.5, 0.8, 10), grid_mode=True)
array([[0.0893    , 0.08919374, 0.08912713, 0.08910291, 0.08912241,
        0.08918551, 0.08929079, 0.08943552, 0.08961588, 0.08982716],
       [0.1276    , 0.126881  , 0.12643032, 0.12626645, 0.12639833,
        0.12682535, 0.12753768, 0.12851697, 0.12973738, 0.13116695],
       [0.09615   , 0.09412557, 0.09285663, 0.09239525, 0.09276657,
        0.09396889, 0.09597453, 0.09873182, 0.10216802, 0.10619312]])

With grid_mode=True (default value), a noise value is computed for every combination of the input value. In this case, the length of the first input is 3, and the length of the second input is 10. The result is an array with shape (3, 10).

If grid_mode=False, all the input must have the same length, and the result has the same shape:

>>> noise.noise2(np.linspace(0.1, 0.3, 30), np.linspace(10, 10.5, 30), grid_mode=False)
array([0.099144  , 0.12303124, 0.14685425, 0.17057388, 0.194133  ,
       0.21745892, 0.24046583, 0.26305716, 0.28512793, 0.30656706,
       0.32725957, 0.34708877, 0.36593838, 0.38369451, 0.40024759,
       0.41549419, 0.42933872, 0.44169499, 0.45248762, 0.46165335,
       0.46914216, 0.4749182 , 0.47896062, 0.48126415, 0.48183955,
       0.4807138 , 0.47793023, 0.47354831, 0.46764335, 0.460306  ])

A seed value can be specified when creating the Noise class or afterward using the seed() function:

>>> noise = vnoise.Noise(4)
>>> noise.seed(5)

This code is available under the MIT license, see LICENSE.

This code is based on Casey Duncan's noise library. The port was done with the help of @tatarize