Neural Interactive Shaping of Parameter Spaces
NISPS lets you teach a neural network how to map a 2D joystick position to a rich set of visual or audio parameters. Move the joystick, shape the outputs you want, and the network learns your preferences in real time.
A small neural network takes your joystick X/Y position as input and produces dozens of output parameters. You teach it by either:
| Touch / Mouse | |
|---|---|
| Drag joystick | Move through parameter space |
| + / − buttons | Positive / negative feedback |
| Bottom bar | Train, Randomize, Clear, Follow mode |
| Keyboard | |
| 1 | Negative feedback (−) |
| 2 | Positive feedback (+) |
| Gamepad (Steam Deck, Xbox, etc.) | |
| Left stick | Joystick control |
| LB (left bumper) | Negative feedback |
| RB (right bumper) | Positive feedback |
| A | Train |
| X | Randomize |
| B | Clear examples |
Toggle to Hands mode in the bottom bar to use your webcam for input.