This release introduces a multi-model tagging system, giving you the choice of two different AI models for auto-tagging your photos.
Upstream:
- ๐ LibrePhotos/Frontend: Multi-Model Tagging System
- You can now choose between two tagging models in Site Settings โ Tagging Model:
- Places365 โ The classic scene recognition model (default)
- SigLIP 2 โ Google's latest vision-language model, running as ONNX. Uses zero-shot classification against a curated vocabulary of 900+ real-world photo tags (people, animals, nature, activities, scenes, weather, and more). Returns the top 10 most relevant tags.
- Tags are stored per model โ switching between models does not delete previously generated tags, and does not require rescanning your library.
- The lightbox sidebar now shows tags with color-coded badges for each model (green for SigLIP 2, blue for Places365).
- Things Albums are dynamically filtered to only show tags from the currently active tagging model.
- Search results use tags from the selected model.
- You can now choose between two tagging models in Site Settings โ Tagging Model:
If you are interested on how I develop this application, you can watch me develop features live on my channel
Sponsoring development:
My work is sponsored by 3 awesome people!
If you like the work I do, then you can support me via GitHub sponsors or via PayPal
