# Convert images to text

> OCR in hypershot turns any text on your screen into copyable text in one drag. It uses macOS Vision or Windows OCR under the hood, and the text lands on your clipboard.

There is a specific kind of tiny rage that comes from looking at a piece of text on your screen and thinking, *I could type that, I am a professional, I will type it.*

You do not need to. You are looking at the text. The machine can read it too.

hypershot has OCR built in, and it works the way you'd hope: you select the text that is on screen, and the text comes out as actual text — on your clipboard, ready to paste.

## The whole move is one drag

From the system tray menu you pick **OCR Capture** (or press the default, `⌘⇧O` on macOS or `Ctrl+Shift+O` on Windows; if you like another combo, assign one in **Settings → Shortcuts**). You click and drag a box around the text. You let go.

*Read a QR code straight off your screen, or drag over text in a screenshot and get text you can edit.*

hypershot runs the recognition, and the recognized text is copied to your clipboard automatically. That is the entire interaction. No "export", no "save to file", no choosing a destination. The text is in your hand, wherever you paste next.

By default it also opens the text in your text editor so you can see what it read (switch that off in **Settings → OCR**). Either way, the clipboard copy happens first.

## It uses the machine it is running on

There is no upload to a server and no waiting on a round trip. On **macOS** it uses the system's **Vision** framework, the same one that powers Live Text. On **Windows** it uses the **Windows OCR** APIs. The recognition happens on your machine, on the pixels you selected.

If the platform does not support native text detection, the OCR menu item and toolbar button simply hide themselves. You are not left staring at a dead button.

## OCR on a capture you already have

You do not have to start from a fresh screen selection. If you already have a capture in the editor — a screenshot of a terminal, a chart with a label, a paragraph you grabbed ten minutes ago — you can run OCR on it there.

Set a crop around the text you want, click the **OCR** button in the toolbar, and the recognized text is copied to your clipboard from the pixels of that capture. The screen you looked at and the text you end up with are the same thing.

## The two habits that make it accurate

OCR is as good as the box you give it, so two small habits matter:

- **Select tight.** Crop close to the text. A box full of empty background and half-adjacent UI gives the recognizer more to guess about.
- **High contrast wins.** Dark text on a light background is the easiest case. If the source is low-contrast or tiny, the results will reflect that — that is the nature of reading pixels, not a bug in the tool.

It handles multiple lines fine, so a whole paragraph or a column of values comes out as one clean block.

## Why this belongs in a screenshot tool

Because the reason you screenshot text is almost never "I want a picture of the text." It is "I want *the text*." The picture was just the container it arrived in.

OCR closes the gap between the screenshot and the thing you actually wanted. You select it, you paste it, and the one thing you were dreading — retyping — never happens.

---

From https://hypernuclear.com/blog/the-ocr-trick/