📦 BoxBox - Enhanced Version

Custom node for ComfyUI for interactive region selection (boxes) with advanced features and critical bug fixes.

🎉 What's New in This Enhanced Version

This is an optimized fork of the original BoxBox with several fixes and new functionalities:

✅ Bug Fixes

1. Image Loading (ComfyUI v0.12.2+)

   • Updated URL construction to use the ComfyUI API helper.
   • Proper encoding for filenames with special characters.
   • Error handling with fallback mechanisms.

2. Backend Stability

   • Uses the standard ComfyUI temp folder.
   • Fixed HTTP responses using web.json_response.
   • Filename sanitization to avoid path conflicts.

3. Precise Coordinate Scaling

   • Fixed duplicate scaling (previously occurring in both frontend and backend).
   • Exact conversion from display space coordinates to original image size.

4. Intermediate Node Support

   • Recursive node chain traversal to find the source image.
   • Compatible with processing nodes (Brightness, Blur, etc.) between LoadImage and BoxSelector.

5. NumPy 2.0 Compatibility

   • Updated server logic with explicit casting (np.uint8, np.float32).
   • Resolved binary incompatibilities in coordinate processing.

6. Adaptive Window Scaling

   • Fixed coordinate misalignment when resizing the browser window.
   • Dynamic calculation of the total scale factor (Server Scale * Browser CSS Scale).

🆕 New Features

1. Aspect Ratio Memory

   • Remembers your last selected proportion via localStorage.
   • Keeps the ratio locked when drawing new selections.

2. Selection Restoration

   • Automatically restores the last selection when reopening the selector.
   • Allows for quick adjustments to existing selections.

3. Improved UX

   • Removed intrusive alerts for small images.
   • Detailed console logging for debugging ([BoxBox]).

---

🎯 Usage

1. Add a LoadImage node.
2. Add a BoxSelector node and connect them.
3. Click the 📦 Select Box button.
4. Draw the region in the popup window.
5. Select an Aspect Ratio if desired and click ✅ Confirm.
6. Use BoxCrop to get the image crop.

🔧 Included Nodes

• 📦 BoxSelector: Interactive selection with ratio locking.
• ✂️ BoxCrop: Crops the image based on selected coordinates.
• 📐 BoxResize: Scales the crop while maintaining fidelity.

---

Documentation synchronized following the node-doc-sync protocol.