media-tag

You can automate image tagging and captioning by batch processing an entire directory using open-source vision-language models like CLIP and BLIP-2 to generate labels and WCAG-style alt text for each file.

Yes, semantic classification supports custom vocabularies for image tagging, allowing you to apply specific labels to media files rather than relying solely on default model predictions.

Per-frame video narration extracts metadata and generates captions for videos by processing individual frames with vision-language models, creating detailed descriptions for media archives.

You can build and query semantic search indexes over media folders by using vision-language models to generate vector embeddings, enabling fast content retrieval across large photo libraries.

Yes, batch tagging media files into a CSV requires PyTorch and OpenCV, along with dependencies like transformers and open_clip_torch, to run the vision-language models for metadata enrichment.

Multi-model orchestration coordinates open-source models like CLIP, SigLIP, BLIP-2, and LLaVA to enrich media metadata, combining different vision-language architectures for comprehensive tagging and captioning.