Drop in your clips and Rushes classifies them using Apple's on-device ML models — B-roll, interview, outdoor, food, screen recording, and more. No manual tagging required.
How It Works
Manual tagging doesn't scale. When you're dealing with hundreds or thousands of clips across multiple shoots, adding metadata by hand is tedious and inconsistent. You end up with an untagged pile of footage that's hard to search and harder to reuse.
When you import footage, Rushes samples frames from each clip and runs them through Apple's Vision framework and a Core ML model. The system analyzes visual content — people, scenes, objects, activities — and assigns relevant tags automatically.
All ML processing runs on your Mac using Apple's Neural Engine (on Apple Silicon) or CPU (on Intel). No frames, thumbnails, or metadata ever leave your machine. There are no cloud APIs, no accounts, and no data collection.
The models require approximately 30MB of memory. On an M1 MacBook Air, processing takes roughly 1-2 seconds per clip.
Tags are used throughout Rushes — for filtering, smart collections, and browsing. Once your footage is tagged, drag clips directly into Final Cut Pro, DaVinci Resolve, Premiere Pro, or any editor. Rushes references your original files, so there's no re-importing or transcoding step.