Converse with Chaos, Sculpt Emergence.
A node-based glitch art application for building image processing pipelines. Chain together transforms, corruption effects, and color manipulations to create unique visual artifacts.
Artifice is a visual programming environment for glitch art. You build processing pipelines by connecting nodes - each node performs a specific operation like loading an image, applying a transform, corrupting data, or saving output. The node-based approach lets you:
- Experiment freely - Rearrange, bypass, or duplicate processing stages without rewriting code
- See results immediately - Real-time preview updates as you modify parameters
- Create complex effects - Chain operations that would be tedious to script manually
- Save and share workflows - Export your node graphs to recreate or share effects
- GLIC-style processing - Segmentation, prediction, and quantization algorithms for structured glitch effects
- Frequency transforms - DCT, FFT, and wavelet decomposition for frequency-domain manipulation
- Data corruption - Bit flipping, byte swapping, data repetition, and structural manipulation
- Pixel sorting - Classic glitch aesthetic with configurable thresholds and sort criteria
- Color space conversion - Work in RGB, HSV, LAB, YCbCr, and other color spaces
- Extensible - Create custom nodes in Python
- Python 3.10 or higher
- pip package manager
# Clone the repository
git clone https://github.com/templeoflum/Artifice.git
cd Artifice
# Create a virtual environment (recommended)
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install
pip install -e .python -m artificeThe workspace opens with a Test Card and Color Space node already connected, so you can start experimenting immediately.
- Add processing nodes - Drag nodes from the palette onto the canvas
- Connect nodes - Click an output port, then click an input port
- Adjust parameters - Select a node to edit its settings in the inspector
- Execute - Press Shift+E or click Execute to process the graph
- Save output - Add an Image Saver node and configure the output path
The Test Card is a procedural calibration image designed to reveal how different effects behave. Each region targets specific operations:
| Pattern | Purpose |
|---|---|
| Color Bars (RGBCMYK) | Test color/channel operations, see how effects treat individual colors |
| Checkerboard (8×8) | Reveal frequency-domain effects (DCT, FFT), compression artifacts |
| Diagonal Lines | Test directional operations like pixel sorting and wavelet transforms |
| Zone Plate | Concentric sine rings that expose aliasing and frequency response |
| Step Wedge | Discrete gray levels to visualize quantization and bit depth reduction |
| Radial Gradient | Test smooth gradients, circular distortions, and vignette effects |
| Fine Checkerboard (16×16) | Higher frequency patterns for detailed frequency analysis |
| Perlin Noise | Test segmentation algorithms and texture-based effects |
| Rainbow Hue Sweep | Full-width spectrum for color space conversion testing |
| Grayscale Gradient | Full tonal range to test contrast, gamma, and tonal response |
When experimenting with a new node, run it on the Test Card first - the structured patterns make it easy to understand what the effect is actually doing.
[Test Card] → [Color Space] → [Bit Flip] → [Pixel Sort]
- Add Bit Flip (Corruption) - set bit position to 7, probability to 1.0
- Add Pixel Sort (Transform) - set mode to "brightness"
- Connect them to the existing Color Space node and execute (Shift+E)
- Now change Color Space to YCbCr or LAB and execute again - observe how the same corruption produces completely different effects depending on color space
- Image Loader - Load PNG, JPG, TIFF, WebP, BMP, GIF
- Image Saver - Save processed images
- Test Card - Procedural calibration image with color bars, gradients, checkerboards, zone plate, and noise patterns
- Color Space - Convert between RGB, HSV, LAB, XYZ, YCbCr, LUV, YIQ
- Channel Split / Merge - Separate and recombine color channels
- Channel Swap - Reorder channels
- Quadtree Segment - Adaptive segmentation by variance, edges, or gradient
- Predict - GLIC-style predictors (Horizontal, Vertical, DC, Paeth, Average, Gradient)
- Quantize - Reduce bit depth with uniform, adaptive, or per-channel modes
- DCT - Discrete Cosine Transform
- FFT - Fast Fourier Transform
- Wavelet - Multi-level wavelet decomposition
- Pixel Sort - Sort pixels by brightness, hue, saturation, or channel value
- Bit Shift / Bit Flip - Bit-level manipulation
- Byte Swap - Byte-level corruption
- XOR Noise - XOR-based noise patterns
- Data Repeat / Drop / Weave / Scramble - Structural data manipulation
- GLIC Pipeline - Combined segmentation → prediction → quantization in one node
from artifice.core.graph import NodeGraph
from artifice.nodes.io.loader import ImageLoaderNode
from artifice.nodes.io.saver import ImageSaverNode
from artifice.nodes.transform.pixelsort import PixelSortNode
graph = NodeGraph()
loader = ImageLoaderNode()
loader.set_parameter("path", "input.png")
sort = PixelSortNode()
sort.set_parameter("sort_by", "brightness")
saver = ImageSaverNode()
saver.set_parameter("path", "output.png")
for node in [loader, sort, saver]:
graph.add_node(node)
graph.connect(loader, "image", sort, "image")
graph.connect(sort, "image", saver, "image")
graph.execute()Artifice/
├── src/artifice/
│ ├── core/ # Node system, graph, data types
│ ├── nodes/ # Node implementations
│ │ ├── io/ # Image loading/saving
│ │ ├── generator/ # Procedural image generation
│ │ ├── color/ # Color space operations
│ │ ├── segmentation/
│ │ ├── prediction/
│ │ ├── quantization/
│ │ ├── transform/ # DCT, FFT, wavelets, pixel sort
│ │ ├── corruption/ # Bit/byte manipulation
│ │ ├── pipeline/ # Combined processing nodes
│ │ └── utility/
│ └── ui/ # Qt-based interface
├── tests/ # Test suite (277 tests)
└── docs/ # Documentation
- Developed and tested on Windows only - macOS and Linux support is untested and may have issues
- Requires Python 3.10 or higher
- Some nodes may be slow on large images without GPU acceleration (planned feature)
- Test suite dialog prompt - Running
pytesttriggers a "Save or Discard" dialog that requires manually clicking "Discard" to continue. Tests will complete normally after dismissing. - High DPI scaling - UI may appear small on high-DPI displays; Qt scaling settings may help
Current focus is on stability and cross-platform compatibility. Planned features:
- Video synthesis - Oscillators, colorizers, keyers, and feedback systems inspired by hardware video synthesizers (LZX, Vidiot, Fairlight CVI)
- Real-time preview - Continuous render loop for live synthesis experimentation
- Video processing - Frame-by-frame processing, temporal effects, frame blending
- Audio processing - Audio codecs, compression artifacts, sonification techniques, spectrogram manipulation
- Audio reactivity - Drive parameters from audio input
- GPU acceleration - CUDA/OpenCL for performance-critical operations
- AI integration - Semantic segmentation, style transfer, learned effects
- CLAUDE.md - Architecture, node development guide, API reference (also serves as context for Claude Code)
- docs/getting-started.md - Detailed tutorial
- docs/node-development.md - Creating custom nodes
Contributions welcome. See CONTRIBUTING.md for guidelines.
# Run tests
pytest
# Run with coverage
pytest --cov=artificeMIT License - see LICENSE
This project is developed using Claude Code, with Claude (Anthropic's AI) handling implementation under human guidance and creative direction. The codebase is structured for continued AI-assisted development - see CLAUDE.md for architecture details and development patterns.
- Developed with Claude Code by Anthropic
- Inspired by GLIC and glitch art research
- Built with PySide6
- Node editor concepts from Blender, Nuke, and TouchDesigner
Converse with Chaos, Sculpt Emergence.