Official website: sololo.xyz
SoloTagger is just a simple Python script with two basic functions:
- Send images and tagging instructions to LLM.
- Receive the output from LLM and save it into a TXT file.
SoloTagger itself does not run the model. It relies on a third-party LLM runtime and has currently been tested with LM Studio on Windows 11. Ollama should also work, but it hasn’t been tested.
If you want to connect to a remote API, you’ll need to make a simple modification to the script to add the API key.
There is currently no GUI. For a simple tool like this, the command line is sufficient.
Clone the repository:
git clone https://github.com/sololo-xyz/SoloTagger.git
Navigate to the SoloTagger directory, then start SoloTagger with:
python SoloTagger.py
So SoloTagger does not include unnecessary third-party libraries just for convenience. It has zero external dependencies. As long as you have Python installed, you can run it.
For installation and usage details, please refer to the documentation:
For prompt editing, please refer to the documentation:
https://sololo.xyz/article/25-solotagger-v012-easier-prompt-editing
Because I often create LoRAs for text-to-image models, I frequently need to generate caption files for images. At the moment, SoloTagger is the main captioning tool I use.
My LoRA works can be browsed and downloaded from my website: sololo.xyz
SoloTagger is currently designed mainly for small LLMs that can run locally, such as the Qwen3.5 series and JoyCaption. My personal recommendations are:
- Qwen3.5 9B GGUF
- JoyCaption beta one GGUF
These two small models provide a good balance of speed and output quality on my laptop.
My laptop specs are: RTX 5060 (8GB VRAM) and 32GB RAM. If your hardware is more powerful, you can choose larger models for better results.