ComfyUI Extension: ComfyUI-Only
A collection of custom nodes for ComfyUI, originally focused on workflow parsing, now expanded to provide advanced file loading features like 'Load Latent (Advanced)' for drag-and-drop latent file uploads and 'WorkflowImageFileLoader' for parsing prompts from image metadata. It also adds a js extension for a better UI experience.
Custom Nodes (0)
README
Read this document in Chinese (็ฎไฝไธญๆ)
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.
๐ Key Features
1. Load Latent (Advanced) โญ Core Feature
- ๐ค Upload from Anywhere: Select
.latent
files from anywhere on your computer using the "Upload Latent" button, eliminating the need to place them in theinput
folder. - โจ Drag and Drop: Upload
.latent
files by simply dragging them from your file manager onto the node. - ๐ Smart Format Compatibility: Automatically detects and supports both standard
pickle
and high-speedsafetensors
formats. - ๐ค Intelligent Structure Parsing: Automatically parses various internal latent structures, whether it's the standard
samples
key, the non-standardlatent_tensor
key, or even a raw tensor. - ๐ข 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
- ๐ Direct File Loading: Load image files directly from your file system.
- ๐ Auto-Parse Metadata: Automatically parses workflow metadata from the image.
- ๐ฏ Smart Prompt Extraction: Intelligently extracts positive and negative prompts.
- ๐ท๏ธ Smart Model Extraction: Extracts the model name from the CheckpointLoaderSimple node.
- โ
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.)
๐ Installation & Setup
Recommended Method: Install via ComfyUI Manager
- Install ComfyUI Manager: If you haven't already, install it by following the official ComfyUI Manager instructions.
- Search for the Node:
- Launch ComfyUI.
- Click the "Manager" button in the sidebar.
- Click "Install Custom Nodes".
- Type
ComfyUI-Only
oreric183
in the search bar.
- Install Node: Find this extension in the search results and click the "Install" button.
- Restart ComfyUI: After installation, restart ComfyUI.
Alternative Method: Manual Installation (Git Clone)
- Navigate to Directory: Open a terminal and navigate to the
custom_nodes
directory within your ComfyUI installation.cd path/to/your/ComfyUI/custom_nodes
- Clone the Repository: Use the
git clone
command to clone this repository.git clone https://github.com/eric183/ComfyUI-Only.git
- 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
- Restart ComfyUI: Restart ComfyUI to load the new nodes.
๐ How to Use
โญ Loading Latent Files (Recommended)
- Add the
Load Latent (Advanced)
node from the menu (under theComfyUI-Only/Latent
category). - Click the
Upload Latent
button to select a.latent
file from your computer. - Alternatively, drag and drop the
.latent
file directly onto the node. - 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
- Add the
WorkflowImageFileLoader
node (under theComfyUI-Only/Image
category). - Select the image file you want to parse from the
image_file
dropdown menu. - The
positive_prompt
,negative_prompt
, andcheckpoint_name
will be automatically extracted.
๐ Project Structure
.
โโโ __init__.py <-- Main entry point, registers nodes and JS extensions
โโโ js/
โ โโโ latent_loader.js <-- Frontend UI logic for the advanced latent loader
โโโ nodes/
โ โโโ image_processing_nodes.py
โ โโโ latent_nodes.py
โโโ LICENSE
โโโ README.md
โ๏ธ Requirements
- Python 3.8+
- ComfyUI
- PyTorch
- safetensors (Core dependency, typically installed with ComfyUI)
๐ Changelog
v2.1.0 - Integration & Simplification (Codename: Orion)
- ๐ Integration: The project has been successfully added to the official ComfyUI Manager list for one-click installation.
- ๐ Simplification: Highly recommended to install via ComfyUI Manager for a streamlined user experience.
- ๐งน Refactor: Removed
requirements.txt
. Dependencies are now managed by ComfyUI or the user as needed, making the project more lightweight. - ๐ 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)
- ๐ฅ New: Released the
Load Latent (Advanced)
node with support for uploading and dragging-and-dropping latent files from any location. - โจ Added: Developed a standalone frontend JS extension for the new node, providing a native-level UI experience.
- ๐ค Enhanced: The loader backend now intelligently handles format compatibility (pickle/safetensors), structure parsing, and dimension processing.
- ๐งน Refactor: Restructured the project, adding a
js
directory andrequirements.txt
. - ๐ Documentation: Overhauled
README.md
to focus on new features and simplify older descriptions.
v1.2.0
- โ Added: Checkpoint model name extraction from CheckpointLoaderSimple nodes.
- โ
All nodes now have a
checkpoint_name
output.
(Older logs have been omitted.)
๐ค Contributing
Issues and Pull Requests are welcome!
๐ License
MIT License - See the LICENSE file for details.