Screenshot Translation

Screenshot translation handles text you can see but cannot copy: select a screen region, run local OCR, and the text is translated automatically. Recognition never leaves your machine.

Features

• One hotkey to start selecting (default Alt+Z)
• The selection snaps to window edges, can be resized from any side, and shows its pixel size live
• Two local OCR engines: Tesseract (default) and Rapid OCR (local plugin)
• Recognized text flows into the mini window or the main window, reusing your existing language and engine settings

How to Use

Capture and translate

1. Press Alt+Z (or click Screenshot in the mini window)
2. Drag over the region, then click Screenshot to confirm or Cancel to abort
3. OCR runs locally; on success the mini window opens and translates the text
4. Edit the source or copy the translation in the result box

Recognize a local image

1. Click Pick image in the mini window, or Pick image in the main window's Text tab
2. Choose a .jpg / .jpeg / .png file
3. The recognized text is sent to that window for translation

Choosing an OCR Engine

Settings → OCR lets you switch engines:

• Tesseract OCR: default, works out of the box
• Rapid OCR: better accuracy, requires installing a local plugin. Use Download and install to fetch it, or Install from local file with the official ZIP package

Install status and error messages are shown on the same page; once installed, switch the engine in the dropdown above.

Tips

Improve Accuracy

• Keep text sharp: avoid blurry, shaky or heavily skewed regions
• Avoid busy backgrounds: solid or low-texture backgrounds work best
• Limit the region: capture only the text you need — large regions are slower and more error-prone
• Watch contrast: dark text on light (or the reverse) recognizes best

Special Cases

• Tables: capture the whole table in one go to preserve row/column relationships
• Multi-column layouts: capture column by column
• Vertical text: limited support; rotate to horizontal first
• Handwriting: printed text only in practice; handwriting accuracy is much lower

Controls

• Alt+Z: start screenshot translation
• Drag any edge of the selection to resize
• Esc or Cancel to abort

Notes

1. OCR text goes through the same translation pipeline as selection translation; adjust language and engine in the translation window
2. Recognition is fully local — images are never uploaded
3. On failure you will see "No text recognized, please capture a region containing text"
4. Decorative fonts, script typefaces and complex formulas recognize poorly
5. Accuracy correlates with image resolution

FAQ

Q: Nothing happens after capturing, or no text is recognized. Why?

A: Usually the region has no recognizable text, too little contrast, or the text is too small or blurry. Zoom in before capturing, or use selection translation / image picking instead.

Q: The recognized text is out of order.

A: Complex layouts (multi-column, mixed graphics) often confuse reading order. Capture region by region, or paste the text into an editor and clean it up before translating.

Q: Rapid OCR says it is not installed.

A: Install the plugin under Settings → OCR. If the download fails, use the official ZIP with Install from local file, then refresh the status and switch the engine.

Q: Does it recognize formulas or code?

A: Mostly plain text. Math and chemical formulas usually come out as garbage, and code indentation/symbols may be lost — use selection translation or paste the source instead.