A comprehensive set of ComfyUI nodes for generating AI-powered autostereograms (Magic Eye images) with procedural texture generation. This project brings the power of DeepStereo into the ComfyUI node-based workflow environment.

Overview

Transform any image into mesmerizing autostereograms using AI-generated depth maps and procedurally created textures. These nodes break down the complex stereogram generation process into modular, interconnected components that can be mixed and matched for endless creative possibilities.

Features

🧠 AI Depth Estimation & Processing

Enhanced MiDaS Integration: State-of-the-art depth estimation with built-in processing
Multiple Models: MiDaS_small, DPT_Large, and DPT_Hybrid
Smart Model Management: Automatic model handling in ComfyUI's controlnet directory
Advanced Depth Controls:
- Contrast and brightness adjustment
- Gamma correction
- Min/Max depth range control
- Optional depth normalization
- Blur for smoothing
- Process width control

🎨 Pattern & Texture Generation

Procedural Texture Methods:
- Content-driven color dots with multiple modes
- Advanced pattern generation with seamless tiling
- Optimized for stereogram use
Pattern Width Control: Matches stereogram separation for optimal results
12 Noise Types:
- RGB: Random color noise
- Grayscale: Monochrome noise
- Colored Dots: Random dot patterns
- Perlin: Enhanced Perlin-like noise
- Voronoi: Cell-based patterns
- Kaleidoscope: Symmetric radial patterns
- Waves: Wave interference patterns
- Cellular: Organic cell-like patterns
- Fractal: Complex self-similar patterns
- Spiral: Spiral-based patterns
- Interference: Wave interference effects
- Crystalline: Geometric crystal-like patterns
Advanced Pattern Controls:
- Frequency control
- Color variation
- Multiple octaves
- Symmetry points
- Seamless tiling
- Pattern width optimization

✨ Enhanced Stereogram Generation

Multiple Algorithms:
- Standard: Basic stereogram generation
- Improved: Better texture distribution
- Layered: Advanced with texture preprocessing
- Central: Enhanced pattern variation
Reference Types:
- Left to right
- Right to left
- Center out (recommended)
Depth Control:
- Up to 50 distinct depth layers
- Layer smoothing control
- Min/Max separation adjustment
Random Dot Stereograms:
- Traditional RDS with depth layer support
- Customizable dot density
- Color control

🎨 Image Effects

Enhanced Image Processing:
- Advanced sharpening
- Multiple blend modes:
  - Multiply/Screen
  - Overlay
  - Darken/Lighten
  - Color Dodge/Burn
  - Hard/Soft Light
  - Difference/Exclusion
- Brightness preservation
- Strength control
- Color blending

🔧 Utility Features

Smart Image Resizing:
- Megapixel-based
- Dimension-based
- Aspect ratio preservation
Texture Transformation:
- Rotation with multiple modes
- Grid pattern generation
- Color inversion
- Size control
Color Management:
- RGB sliders
- Preset colors
- Hex input
- Format options

Installation

Clone this repository into your ComfyUI custom_nodes directory:

cd /path/to/ComfyUI/custom_nodes/
git clone https://github.com/yourusername/ComfyUI-DeepStereo-Nodes.git

Install the required dependencies:

cd ComfyUI-DeepStereo-Nodes
pip install -r requirements.txt

Restart ComfyUI

Note: This extension requires PyTorch, which should already be installed with ComfyUI. If you encounter torch-related issues, ensure your ComfyUI installation includes proper CUDA support for GPU acceleration.

Quick Start

Recommended Workflow

Depth Map Creation:
- Load Image → MiDaS Depth Estimator
- Use DPT_Large model for best quality
- Adjust contrast, brightness, and depth range
- Enable depth normalization for optimal results
Texture Generation (choose one):
- Random Noise Generator:
  - Set pattern_width to match max_separation (default 100)
  - Try kaleidoscope, crystalline, or interference patterns
  - Enable seamless tiling
- Procedural Texture Generator:
  - Use content-driven patterns
  - Apply texture transformation if needed
Stereogram Creation:
- Use Stereogram Generator with:
  - Algorithm: "central"
  - Reference type: "center_out"
  - Depth layers: 50
  - Layer smoothing: 1.0
- Or try Random Dot Stereogram Generator for classic effects

Advanced Techniques

Pattern Enhancement:
- Use Image Effects for sharper, more defined patterns
- Try different blend modes for unique effects
- Adjust pattern frequency and color variation
Depth Optimization:
- Fine-tune depth layers for desired 3D effect
- Use gamma correction for depth emphasis
- Apply subtle blur for smoother transitions
Texture Tips:
- Avoid blurry or overly complex patterns
- Use high contrast for better depth perception
- Keep pattern size consistent with separation values

Node Categories

DeepStereo/Depth

MiDaS Depth Estimator: AI-powered depth map generation with integrated processing

DeepStereo/Texture

Procedural Texture Generator: Multi-method texture creation
Random Noise Generator: Advanced pattern generation with 12 noise types
Texture Transformer: Rotation, grid, and transform effects

DeepStereo/Effects

Image Effects: Advanced image processing and blending

DeepStereo/Generation

Stereogram Generator: Create sophisticated autostereograms
- Multiple algorithms and reference types
- Depth layer control
- Advanced smoothing
Random Dot Stereogram: Enhanced RDS generation
- Depth layer support
- Advanced controls
- Color customization

DeepStereo/Utility

Image Resizer: Smart dimension control
Color Picker: Advanced color selection

Requirements

ComfyUI (with PyTorch)
OpenCV (opencv-python>=4.5.0)
NumPy (numpy>=1.19.0)
tqdm (tqdm>=4.50.0)
PIL/Pillow (usually included with ComfyUI)

Credits and Inspiration

This ComfyUI node collection is based on my DeepStereo standalone utility. DeepStereo is an integration of AI depth estimation with procedural texture generation for easy and fun autostereogram creation.

Original DeepStereo Features Translated to Nodes:

Complete CLI functionality broken into modular nodes
All texture generation methods (M1-M4) available
Multiple stereogram algorithms implemented
Advanced texture transformation and blending capabilities

If you prefer command-line usage or need batch processing capabilities, check out the original DeepStereo repository.

License

This project is open source. See LICENSE file for details.

Contributing

Issues and pull requests welcome! This project aims to make autostereogram generation accessible to everyone through ComfyUI's intuitive interface.

Create mesmerizing Magic Eye images with the power of AI and procedural generation!