allenv0/SCM

What it solves

It provides a private, local-first search engine for your personal photo and video library on macOS. It solves the problem of finding specific moments or information buried in large collections of media by using AI to understand visual content, spoken dialogue, and on-screen text without ever uploading your files to the cloud.

How it works

The application runs entirely on your Mac, using local machine learning models to index your media. It uses vision models (like CLIP or SigLIP) to create embeddings for images and video frames, Whisper for transcribing spoken dialogue, and Tesseract for OCR. When you search, the app performs semantic matching against these local indexes. You can also opt-in to a local LLM (via llama.cpp) to chat with your media library, allowing you to ask questions and receive cited answers based on your own files.

Who it’s for

It is designed for macOS users who want to organize and search their personal media collections while maintaining strict data privacy. It is particularly useful for those with large archives of videos and photos who need to find specific scenes, spoken lines, or visible text without relying on third-party cloud services.

Highlights

• Local-first privacy: No accounts, no cloud, and no data uploads; all inference runs on your machine.
• Deep video search: Segments videos into scenes, allowing you to jump to the exact moment a specific event occurs.
• Multimodal retrieval: Supports searching by visual meaning, spoken dialogue, visible text (OCR), and filenames.
• Smart organization: Automatically detects screenshots and email addresses within images, with features to save custom search queries as persistent tabs.
• Extensible AI: Switchable vision models and optional local LLM integration for conversational querying of your library.