>>17767458You can create pokefusions with machine learning if you want. What you do is you train an neural net to associate a vector with a certain pokemon sprite and then you plug in different vector values to get fusions.
The vector [ 1 0 0 0 ]^T would be the first pokemon, the vector [ 0 1 0 0 ]^T would be the second ect. This one could have 4 Pokemon as training data. The neural net would take a vector and try to generate the correct Pokemon. You train it by comparing the generated image with the correct image. After you have it trained well enough, you can plug in vector values like [1/sqrt(2) 0 1/sqrt(2) 0 ] to get fusions.
I am pretty sure it doesn't work too well. I watched some video about it probably 4 years ago.
