Video Text Select

Text on screen in a video is trapped. A command in a terminal recording, an error message in a bug report, a URL on a conference slide, a config snippet in a tutorial, a book title on a shelf in the background — you can see it, but you cannot copy it. So you pause, squint, and retype it by hand, and get a character wrong, and try again. macOS gives you Live Text on images. Your browser gives you nothing on video. Video Text Select closes that gap. HOW IT WORKS Press Cmd+Shift+S (Ctrl+Shift+S on Windows and Linux) while a video is playing. The video pauses, the frame freezes, and a moment later every word in the picture becomes selectable text. Drag across it exactly as you would on a web page and press Cmd+C. Or click "Copy all" to take everything at once. Press Esc to dismiss. Playback resumes only if the extension was what paused it — if you had already paused the video yourself, it stays paused. The selection is real browser text selection, not a custom imitation. That means the I-beam cursor, the highlight, double-click to select a word, right-click to copy, and Cmd+C all behave exactly as you already expect, because they are your browser's own features working on text laid over the frame. EVERYTHING RUNS ON YOUR DEVICE The text recognition engine and its English language model are bundled inside the extension. There is no server, no API key, no account, and no network request at any point. Your video frames are processed in memory on your own machine and are never uploaded, stored, or seen by anyone — including us. This is also why it works offline, and why it stays fast: recognising a frame takes a fraction of a second. IT RUNS ONLY WHEN YOU ASK IT TO The extension has no access to any page until you press its shortcut, and then only to the single tab you are looking at. It requests no permission to read your browsing history, your other tabs, or any specific website. It declares no script that runs when a page loads — nothing happens anywhere until you invoke it. WHERE IT WORKS Ordinary HTML5 video: YouTube, Vimeo, self-hosted players, videos embedded in other pages, conference talks, screen recordings, and meeting recordings. Also canvas-based players. WHERE IT DOES NOT DRM-protected streaming services — Netflix, Disney+, Prime Video — deliberately hand back a black frame to anything trying to read their pixels. That is by design and cannot be worked around. Rather than showing you an empty overlay and letting you wonder, the extension detects it and tells you plainly. Recognition is English only for now. Accuracy is very good on the kind of text you actually want to copy — slides, code, terminals, subtitles, UI. Stylised titles and very low-resolution text are harder, as they are for any recognition engine. OPEN SOURCE The full source is at https://github.com/hpfs74/copy-text-video — including the tests, the build, and every line of what runs on your machine.