Repolode
Submit
allenv0

allenv0/SCM

Local-first macOS app that lets you search photos and video frames in any folder by describing them in plain language, using on-device AI models.

Open-source alternative to:Google PhotosApple PhotosAdobe Premiere Pro

View on GitHubWebsite
Stars
488
Forks
38
Open issues
1
Last commit
3 hours ago

What is it?

SCM (Screen Memories) is a macOS desktop app for searching the photos and videos in your folders. You type a description such as a place, an object or a phrase, and the app returns matching whole files, the exact moment inside a video, or images containing matching on-screen text. Search can run on vision embeddings from CLIP, SigLIP-2 or EmbeddingGemma-2, while OCR via Tesseract and spoken-word search via Whisper cover text and dialogue.

The app is local-first. Media is copied into an app-managed library under Application Support, and model weights are downloaded once, after which indexing and search run offline. There are no accounts, telemetry or uploads. An optional LLM chat mode uses a llama.cpp sidecar to answer questions from the extracted dialogue, OCR text and filenames, with citations back to the source.

The project is built with TypeScript, React and Electron, uses Bun as its package manager, and ships as a DMG or ZIP via electron-builder or a Homebrew cask. It includes watched-folder import, content-hash deduplication, saved search tabs and a Video Search pipeline that samples scenes at configurable density.

Key features

  • 01Searches photos and videos by meaning using local vision embedding models.
  • 02Jumps to the matching timecode inside a video scene.
  • 03Finds literal text in images via Tesseract OCR with multiple language packs.
  • 04Searches exact spoken lines in videos using Whisper transcripts.
  • 05Runs an optional local LLM chat that answers from extracted evidence with citations.
  • 06Auto-imports watched folders and deduplicates files by SHA-256 content hash.

Use cases

  • Finding a specific photo in a large personal library by describing it.
  • Locating the moment a phrase is spoken in recorded video or meetings.
  • Searching screenshots and scanned images for visible text or email addresses.
  • Keeping a private, offline media archive on a Mac without uploading files to the cloud.

Topics

#multimodal-search#local-ai#macos#electron#video-search#ocr#clip#whisper#embeddinggemma-2#siglip-2

FAQ

What is SCM?
Local-first macOS app that lets you search photos and video frames in any folder by describing them in plain language, using on-device AI models. SCM (Screen Memories) is a macOS desktop app for searching the photos and videos in your folders. You type a description such as a place, an object or a phrase, and the app returns matching whole files, the exact moment inside a video, or images containing matching on-screen text. Search can run on vision embeddings from CLIP, SigLIP-2 or EmbeddingGemma-2, while OCR via Tesseract and spoken-word search via Whisper cover text and dialogue.
Is SCM free to use? What license does it have?
SCM is open source under the MIT license. Check the license terms before using it in commercial products.
Is SCM still maintained?
Actively maintained. The last commit to SCM was on Oct 18, 2026.
What can I use SCM for?
Finding a specific photo in a large personal library by describing it. Locating the moment a phrase is spoken in recorded video or meetings. Searching screenshots and scanned images for visible text or email addresses. Keeping a private, offline media archive on a Mac without uploading files to the cloud.
Where can I get SCM?
The source code and installation instructions are on GitHub: https://github.com/allenv0/SCM

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects