ComfyUI_SSStereoscope

ComfyUI SS Stereoscope

Transform any 2D image into immersive stereoscopic 3D content in ComfyUI.

The SideBySide Stereoscope Node is a cutting-edge 2D-to-3D conversion tool designed for ComfyUI. Powered by Depth-Anything (V2 & V3), it automatically creates high-quality depth maps and generates side-by-side (SBS) 3D images and videos perfect for VR headsets and 3D displays.

✨ Features

• 🧠 AI-Powered Depth - Uses Depth-Anything (V2 & V3) for automatic, high-quality depth estimation.
• 👓 Stereoscopic Modes - Supports Parallel and Cross-eyed viewing techniques.
• 🎥 Video Support - Full workflow for converting videos to 3D (upload, process, combine).
• 🎛️ Fine Control - Adjust depth scale, blur radius, and depth inversion.
• ⚡ Integrated Workflow - No external tools needed; handles everything within ComfyUI.
• 🚀 Performance Mode - Optional "HighSodium Optimization" for up to 4x faster processing.

📸 Demo

🚀 Quick Start

Prerequisites

• ComfyUI installed
• Python >= 3.8
• FFmpeg (Recommended for video support)
  • The node attempts to use your system's FFmpeg first.
  • If missing, it will fallback to the imageio-ffmpeg binary (included in dependencies).
  • Windows Users: For best performance, install FFmpeg manually from ffmpeg.org and ensure it is added to your System PATH.

Installation

1. Navigate to your custom nodes directory:

   cd ComfyUI/custom_nodes
   

2. Clone the repository:

   git clone https://github.com/SamSeenX/ComfyUI_SSStereoscope.git
   

3. Install dependencies:

   pip install -r ComfyUI_SSStereoscope/requirements.txt
   

4. Restart ComfyUI.

(Note: The AI model will download automatically on first use.)

📖 Documentation

Core Nodes

• 👀 SBS V2: The main node. Input a 2D image, get a 3D SBS image + depth map.
  • depth_scale: Intensity of the 3D effect.
  • mode: Cross-eyed vs Parallel.
  • highsodium_optimization: Enable vectorized algorithm for faster processing (may produce slightly different results).
• 👀 SBS V2.1 (External Depth): New in v2.1! Uses the fast V2 engine (HighSodium) but accepts your custom Depth Maps.
  • Fixes "reducing" artifacts found in the legacy node.
  • Resolution-Relative Scaling: depth_scale (0-100) now maps to 0-20% of image width. This gives you fine-grained control while preventing output-breaking values.
    • Scale 10 = 2% width separation.
    • Scale 50 = 10% width separation.
    • Scale 100 = 20% width separation (Max).
• 👀 SBS Video Uploader: Converts input video to image sequence.
• 👀 SBS Video Combiner: Merges processed frames back into a video.

Workflow Example

1. Video Input: Use SBS Video Uploader to load a video.
2. Process: Connect frames to SBS V2 node.
3. Output: Connect results to SBS Video Combiner to save the 3D video.

🏗️ Project Structure

ComfyUI_SSStereoscope/
├── sbs_v2.py            # Main logic
├── requirements.txt     # Dependencies
├── video_utils.py       # Video processing helpers
└── README.md

🤝 Contributing

Contributions are welcome!

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

Contributors

• SamSeen - Original Author
• HighSodium - Performance optimization (PR #16)

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

☕ Support

If you find this project useful, please consider supporting me:

• ⭐ Starring this repository
• 🐛 Reporting issues
• ☕ Buy me a coffee

---

Created with ❤️ by SamSeen