SCM launches offline AI search tool for photos and videos on macOS

A new macOS application called SCM now lets users search their local photos and video frames using plain language, running entirely offline on Apple Silicon Macs. The tool is free and requires no accounts, cloud uploads, or telemetry, which means creative professionals can find visual assets and spoken dialogue without sending client files to external servers.

What the tool does

SCM performs all inference locally using a combination of AI models. A vision model handles visual matching, OCR extracts text from screenshots and video frames, and Whisper transcribes spoken dialogue for exact phrase search. The application supports four switchable models via ONNX Runtime, with a default CLIP model that requires a one-time download of roughly 435MB.

Users can also opt into a llama.cpp sidecar for local chat querying over extracted evidence. The software handles automatic library maintenance, watched folders, and deduplication through SHA-256 content hashing.

Practical uses for working professionals

Marketers and founders are the primary audience, but the tool has clear applications across several fields. Real estate agents can search property walkthrough videos for specific room features. Event planners can locate candid shots by describing scenes. Writers and content creators can pull dialogue from interview footage without scrubbing through timelines.

The search covers three distinct modes: visual match for finding objects or scenes, OCR text search for pulling words from screenshots and video frames, and exact dialogue retrieval for transcribed speech.

Setup and requirements

Installation takes minutes through Homebrew with the command brew install --cask allenv0/scm/scm. After installing, users import folders containing photos and videos, then type plain-language search queries. The application requires Apple Silicon and macOS 12 or higher.

Why this matters for creatives, marketers, and event professionals

Client assets often contain sensitive material that cannot be uploaded to cloud services. SCM removes that friction by keeping everything on-device. A marketing manager searching years of campaign footage for a specific product shot, or a hospitality team locating event photos by describing table layouts, can now do so in seconds without compliance headaches. For professionals already building local AI workflows, the opt-in chat sidecar adds a layer of querying that sits on top of extracted evidence rather than requiring manual file-by-file review.