I want to discuss this new approach to procedural generators I have found on Github.
Despite its attention-seeking name it has not much to to with the mathematics of quantum mechanics.
But, I find the results astonishing none the less.
As far as I understand the algorithm it basically works like this:
- More like machine learning and unlike peril noise based approaches, it takes a sample input data set.
- It decomposes the input into elements (currently using a uniform grid).
- It calculates the probability of one type of element neighboring another.
- It fills a new output set randomly with elements from the input in a fashion that keeps the neighboring probabilities intact.
- You get something similar to the input but it is randomly generated and can be far bigger.
I could image this being an add-on for blender, the question is: Would anybody care?