Read this document in Chinese (简体中文)

ComfyUI-Only

A collection of custom nodes for ComfyUI, initially focused on workflow parsing, now expanded to provide advanced file loading features designed to enhance your workflow efficiency and experience.

🌟 Key Features

1. Load Latent (Advanced) ⭐ Core Feature

• 📤 Upload from Anywhere: Select .latent files from anywhere on your computer using the "Upload Latent" button, eliminating the need to place them in the input folder.
• ✨ Drag and Drop: Upload .latent files by simply dragging them from your file manager onto the node.
• 🔄 Smart Format Compatibility: Automatically detects and supports both standard pickle and high-speed safetensors formats.
• 🤖 Intelligent Structure Parsing: Automatically parses various internal latent structures, whether it's the standard samples key, the non-standard latent_tensor key, or even a raw tensor.
• 🔢 Smart Dimension Handling: Automatically processes 3D (image), 4D (batched images), and 5D (video) latent tensors to ensure compatibility with downstream nodes like VAEDecode.

2. WorkflowImageFileLoader

• 📁 Direct File Loading: Load image files directly from your file system.
• 🔍 Auto-Parse Metadata: Automatically parses workflow metadata from the image.
• 🎯 Smart Prompt Extraction: Intelligently extracts positive and negative prompts.
• 🏷️ Smart Model Extraction: Extracts the model name from the CheckpointLoaderSimple node.
• ✅ Solves a Core Problem: Addresses the issue where the native IMAGE type does not contain metadata, making workflow reuse difficult.

---

(Older node descriptions have been collapsed for brevity.)

🚀 Installation & Setup

Recommended Method: Install via ComfyUI Manager

1. Install ComfyUI Manager: If you haven't already, install it by following the official ComfyUI Manager instructions.
2. Search for the Node:
   • Launch ComfyUI.
   • Click the "Manager" button in the sidebar.
   • Click "Install Custom Nodes".
   • Type ComfyUI-Only or eric183 in the search bar.
3. Install Node: Find this extension in the search results and click the "Install" button.
4. Restart ComfyUI: After installation, restart ComfyUI.

Alternative Method: Manual Installation (Git Clone)

1. Navigate to Directory: Open a terminal and navigate to the custom_nodes directory within your ComfyUI installation.

   cd path/to/your/ComfyUI/custom_nodes
   

2. Clone the Repository: Use the git clone command to clone this repository.

   git clone https://github.com/eric183/ComfyUI-Only.git
   

3. Check Dependencies: Ensure that the safetensors library is installed in your Python environment. ComfyUI usually includes it, but if you encounter import errors, install it manually:

   pip install safetensors
   

4. Restart ComfyUI: Restart ComfyUI to load the new nodes.

📖 How to Use

⭐ Loading Latent Files (Recommended)

1. Add the Load Latent (Advanced) node from the menu (under the ComfyUI-Only/Latent category).
2. Click the Upload Latent button to select a .latent file from your computer.
3. Alternatively, drag and drop the .latent file directly onto the node.
4. The node will automatically handle the upload, parsing, and formatting, then connect the LATENT output to the next node (e.g., VAEDecode).

Parsing Workflows from Images

1. Add the WorkflowImageFileLoader node (under the ComfyUI-Only/Image category).
2. Select the image file you want to parse from the image_file dropdown menu.
3. The positive_prompt, negative_prompt, and checkpoint_name will be automatically extracted.

📁 Project Structure

.
├── __init__.py           <-- Main entry point, registers nodes and JS extensions
├── js/
│   └── latent_loader.js  <-- Frontend UI logic for the advanced latent loader
├── nodes/
│   ├── image_processing_nodes.py
│   └── latent_nodes.py
├── LICENSE
└── README.md

⚙️ Requirements

• Python 3.8+
• ComfyUI
• PyTorch
• safetensors (Core dependency, typically installed with ComfyUI)

📝 Changelog

v2.1.0 - Integration & Simplification (Codename: Orion)

• 🎉 Integration: The project has been successfully added to the official ComfyUI Manager list for one-click installation.
• 🚀 Simplification: Highly recommended to install via ComfyUI Manager for a streamlined user experience.
• 🧹 Refactor: Removed requirements.txt. Dependencies are now managed by ComfyUI or the user as needed, making the project more lightweight.
• 📚 Documentation: Completely rewrote README.md with updated installation instructions, project structure, and dependency information for clarity and accuracy.

v2.0.0 - Advanced Loader Edition (Codename: Phoenix)

• 🔥 New: Released the Load Latent (Advanced) node with support for uploading and dragging-and-dropping latent files from any location.
• ✨ Added: Developed a standalone frontend JS extension for the new node, providing a native-level UI experience.
• 🤖 Enhanced: The loader backend now intelligently handles format compatibility (pickle/safetensors), structure parsing, and dimension processing.
• 🧹 Refactor: Restructured the project, adding a js directory and requirements.txt.
• 📚 Documentation: Overhauled README.md to focus on new features and simplify older descriptions.

v1.2.0

• ✅ Added: Checkpoint model name extraction from CheckpointLoaderSimple nodes.
• ✅ All nodes now have a checkpoint_name output.

(Older logs have been omitted.)

🤝 Contributing

Issues and Pull Requests are welcome!

📄 License

MIT License - See the LICENSE file for details.