Generate Gradients and Patterns using Neural Networks

This is the source code for a section of my website. You’re welcome to snoop around to see how it works, maybe even make a pull request! (I’ve been wanting to add an interesting UI!)

How it works

Refresh the page to regenerate a random Neural Network and start drawing. It works by feeding the X and Y coordinate into the neural network, and using the three outputs for the red, green and blue value at that pixel. Different kinds of Networks produce different aesthetics.

Plans for the future

Mostly UI stuff, you don’t have any control over things otherwise.

  • A way to redraw the image (click it? a keybind?)
  • Network customisation
    • Number of hidden layers, and nodes per layer should be changable
    • A way to change the inputs, like adding a Random value for a dithery effect
    • idk stuff like that
  • Rendering the image at custom resolutions.


View Github