ComfyUI WD 1.4 Tagger

If you cannot use the automatic downloader, you can manually provide models.

1. Create a models folder in the same directory as wd14tagger.py.
2. Obtain the model URLs from the list provided in pysssss.json.
3. Download model.onnx and rename it to match the model name (e.g., wd-v1-4-convnext-tagger-v2.onnx).
4. Download selected_tags.csv and rename it to match the model name (e.g., wd-v1-4-convnext-tagger-v2.csv).

You can perform a quick interrogation on any node that displays an image (such as LoadImage, SaveImage, or PreviewImage).

1. Right-click on the node (or on a specific image if the node displays multiple).
2. Select WD14 Tagger from the context menu.

Settings for this quick interrogation feature are managed in the settings section of pysssss.json.

To interrogate images for booru tags, add the node via the menu: image -> WD14Tagger|pysssss.

Models are automatically downloaded at runtime if they are missing. The node supports tagging and outputting multiple batched inputs.

Node Parameters:

• model: The interrogation model to use (e.g., MOAT or ConvNextV2).
• threshold: The minimum score for a tag to be considered valid.
• character_threshold: The minimum score for a character tag to be considered valid.
• exclude_tags: A comma-separated list of tags to exclude from the results.

Follow these steps to install the extension into your ComfyUI setup:

1. Clone the repository into your custom_nodes folder:

   git clone https://github.com/pythongosssss/ComfyUI-WD14-Tagger

2. Navigate to the newly created extension folder in your terminal:

   cd custom_nodes/ComfyUI-WD14-Tagger

3. Install the required Python dependencies based on your installation type:

For Windows Standalone (embedded python):

../../../python_embeded/python.exe -s -m pip install -r requirements.txt

For Manual/Non-Windows installations:

pip install -r requirements.txt

git clone https://github.com/pythongosssss/ComfyUI-WD14-Tagger
cd custom_nodes/ComfyUI-WD14-Tagger
# For Windows Standalone:
../../../python_embeded/python.exe -s -m pip install -r requirements.txt
# For Manual:
pip install -r requirements.txt

The extension requires onnxruntime to function.

• CPU: Interrogation is fast on CPU using standard onnxruntime (included in requirements.txt).
• GPU: You can use onnxruntime-gpu for GPU acceleration, though support for this specific package is not provided if installation issues occur.