Getting started

Install

  1. Download OpenClip_0.2.0.dmg, open it, and drag OpenClip to Applications.
  2. OpenClip is ad-hoc signed and not notarised. The first time, right-click the app and choose Open, then confirm. (Or run xattr -dr com.apple.quarantine /Applications/OpenClip.app in Terminal.)

You need macOS 13 or later. OpenClip is tested on Apple silicon.

Import a video

Drop a long 16:9 video onto the window. OpenClip transcribes it on your Mac with whisper.cpp (using the Metal GPU, with word-level timestamps) while Apple Vision finds faces in parallel.

Choose a model

ModelUse it when
Fast (base.en)You want the quickest pass.
Balanced (small.en)A good default for clear English speech.
Best (large-v3-turbo)You want the most accurate transcript.

The model is downloaded once from Hugging Face, only when you ask for it. After that, everything works offline. As a measured data point, a 75-minute 2 GB 720p sermon was transcribed and its faces found in about 8.5 minutes on an M-series Mac (transcription near 9× realtime on small.en).

Pick the clips

Ask Claude to find clips, or use the built-in Quick pick. Then preview, trim and tick the ones you want, and export them together.

Build from source

Install Rust (stable) and build from the repository at github.com/MarshallNickolauson/OpenClip. See its README for the current build steps.