Show HN: AI search for every photo and every frame of video on macOS
An open-source, local-first macOS app called SCM provides AI-powered natural language search across local photos, video frames, OCR text, and audio dialogue.
summary is enough The project description outlines a comprehensive local-first macOS media search tool, while the comment thread offers sharp technical critiques on framework choices.
summary
SCM is a local-first macOS application that enables natural language semantic search across photos and individual video frames without cloud uploads. It combines vision models like CLIP and SigLIP with Tesseract OCR, Whisper for dialogue transcription, and an optional local LLM sidecar for chat. The project runs locally via Electron and Bun, offering features like deduplication, scene segmentation, and background indexing. Commenters debate its technology choices, such as using Tesseract over native Apple APIs, and discuss the implications of AI-assisted software generation.
key points
- Performs local-first semantic search across photos and video frames using swappable vision models like CLIP and SigLIP.
- Extracts text and dialogue via Tesseract OCR and Whisper, enabling exact-phrase search and optional local LLM chat.
- Commenters point out that leveraging Apple's native Vision framework would provide superior OCR speed and accuracy compared to Tesseract.
who should read it
macOS developers and power users interested in local-first AI tools and media asset management.
discussion positive
consensus
Commenters find the local-first macOS media search concept appealing and useful for video editing workflows.
pushback
Technical users criticize the choice of Tesseract over Apple's native Vision framework, attributing it to suboptimal patterns in LLM-assisted code generation.
Since this is for the mac you really should be using apple's vision framework for OCR. It smokes tesseract in both speed and accuracy.
Edit: I'm curious which LLM was used to generate the code. I fed the title of your post to claude/deepseek/qwen/codex asking to recommend a stack for this project, expecting to frown thinking that they still recommend tesseract. However, I found that they all recommend apple's vision framework. In fact the latest model to recommend Tesseract is gpt-4.1.
Points out that macOS provides a native Vision framework that significantly outperforms Tesseract in both speed and accuracy.
Generated by an LLM from the article and thread. Check the sources before relying on it.