| English | README.md |
| 简体中文 | README_Zh-CN.md |
| 繁體中文 | README_Zh-TW.md |
Single Image
Batch Process
Batch from Folder
Supports SD-WebUI-forge and Neo branches.
Automatic model download is not supported for the time being. Please manually download the model files and place them in the corresponding model directory.
Place model files in SD-WebUI-Forge/model/tagger_models; models must have the file suffix .onnx!
The model folder needs to be automatically generated by running the plugin.
- Open WebUI and navigate to
Extensions->Install from URL. - Enter the URL:
https://github.com/Tao256/sd-forge-wd14-tagger.git. - Click
Install, and restart WebUI after installation is complete.
- Go to the Releases page to download the latest version of the compressed package.
- Place the unzipped folder into the
SD-WebUI-Forge/extensions/directory. - Restart WebUI, and the plugin will load automatically.
| Model | Confidence Threshold | Model Download | Description |
|---|---|---|---|
| wd-convnext-tagger-v3 | 0.3–0.5 | HuggingFace Modelscope |
Pure convolutional architecture with fast inference and moderate VRAM usage, suitable for batch processing |
| wd-eva02-large-tagger-v3 | 0.4–0.6 | HuggingFace Modelscope |
Better at capturing the overall atmosphere and style of images, with strong tag relevance |
| wd-swinv2-tagger-v3 | 0.35–0.55 | HuggingFace Modelscope |
High recognition accuracy for typical anime features such as hair color, eye color, and clothing style |
| wd-vit-tagger-v3 | 0.3–0.45 | HuggingFace Modelscope |
Strong semantic relevance between generated tags, suitable for direct use as prompt words |
| wd-vit-large-tagger-v3 | 0.35–0.5 | HuggingFace Modelscope |
ViT-Large has a larger parameter count and stronger ability to recognize complex compositions and detailed features |
- Project Homepage: https://github.com/Tao256/sd-forge-wd14-tagger
- Stable Diffusion WebUI Neo: https://github.com/lllyasviel/stable-diffusion-webui-forge
- WD14 Tagger: https://github.com/picobyte/stable-diffusion-webui-wd14-tagger