An OCR vocabulary app for everywhere words appear

Subtitles, screenshots, protected readers, scanned course materials, the text inside an app's own interface: useful words constantly show up in places you can't select. A vocabulary tool that only accepts typed or copied text quietly excludes a huge share of your real reading.

An OCR vocabulary app closes that gap by working from what's visible rather than what's selectable. Capecho uses macOS's built-in on-device text recognition, so any word you can see becomes a word you can save, whether it lives in a video frame, an image, a locked PDF, or a panel that won't let you highlight anything.

Video subtitles are the clearest example: a phrase you'd like to keep flashes by and is gone before you can pause and copy it. Save words from videos and the friction disappears, because you're reading pixels, not a transcript you don't have.

Scanned PDFs and image-based course packs are the same story. They look like text but behave like pictures, so highlighting does nothing. The ability to save words from PDFs and save words from images is exactly where an OCR approach earns its place, turning the documents you actually study into a source you can collect from.

Capecho doesn't ship its own recognition engine or send your screen to a server to be read. It calls the operating system's own text recognition, the same technology behind Live Text, and the work happens on your device. That keeps capture fast and keeps the contents of your screen out of the cloud.

macOS gates that capability behind a permission it labels Screen Recording, but the name describes the OS mechanism, not what Capecho does. Capecho never records or streams your screen. It reads a single frame at the moment you press the shortcut and nothing else.

There's nothing running in the background and no continuous scanning. Recognition fires only at the instant you trigger the shortcut and returns just the text it reads — the screen image itself never reaches Capecho. What it holds on to is the word and its sentence, not a picture of your screen.

If you'd prefer not to grant the screen permission at all, a copy-paste mode reads only the selection you've copied, again only after you press the shortcut, never by watching the clipboard. It's a deliberate alternative, not a degraded fallback.

Recognition is good, but you're the final check. Every capture opens a preview you can edit: correct a mis-read character, adjust the sentence Capecho grabbed, or remove anything sensitive that happened to be on screen. Only the word and context you approve are stored.

That editable step is what makes powerful capture trustworthy. You're curating your own library rather than accepting whatever the camera saw, and the captured word stays fixed afterward while the context sentence and its gloss remain yours to refine.

A captured word doesn't just sit in a list. It comes with a free word explanation built from the word alone (meaning, part of speech, distinct senses, pronunciation, and a macOS Dictionary handoff), and it returns as an FSRS spaced-repetition card fronted by your own sentence, all of it on your Mac today.

English is the first quality-validated target, but you can capture and save other languages now, and you can export your context-rich cards to Anki or CSV whenever you like. A phone companion for reviewing on the go is coming, so the words you capture from a video or a PDF at your desk will be able to return in the spare minutes of your day.