CPPN Art Generator

Generate abstract morphing animations using Compositional Pattern Producing Networks (CPPNs).

What is CPPN?

CPPNs create artistic images by passing coordinate information through neural networks with various activation functions. Think of it as evolution meets neural networks meets generative art.

Quick Start

# Install dependencies
pip install -r requirements.txt

# Generate a single image
python cppn.py

# Generate an animated video
python render_video.py --n_frames 120 --output my_video.mp4

Features

• Hardware Accelerated: Automatic detection and use of Apple MPS, CUDA, or CPU
• Batch Processing: Optimized video generation with efficient GPU utilization
• Color Matching: Match generated art to reference color palettes
• Smooth Animations: Spline interpolation for seamless video transitions
• High Quality Output: 16-bit image support for smooth gradients

Basic Usage

Generate Images

from cppn import generate_image

# Simple generation
generate_image(width=512, height=512, seed=42, output_path="art.png")

# With color matching
generate_image(
    width=512,
    height=512,
    color_reference="reference.jpg",
    color_match_strength=0.8
)

Generate Videos

# Full HD video with 240 frames
python render_video.py --n_frames 240 --width 1920 --height 1080 --fps 30

# With color matching
python render_video.py --n_frames 120 --color_reference colors.jpg --color_match_strength 0.7

Parameters

Parameter	Description	Default
net_size	Network complexity (higher = more detail)	32
h_size	Latent vector size	32
scaling	Zoom level (higher = more zoomed out)	10.0
seed	Random seed for reproducibility	None
num_layers	Number of hidden layers	3

Requirements

• Python 3.7+
• NumPy
• Pillow
• PyTorch (optional, for GPU acceleration)
• SciPy (for video generation)
• FFmpeg (for video encoding)

Credits

Inspired by hardmaru's CPPN implementation. Modernized with PyTorch and optimized for 2025.