pandite

Text in images

pandite reads the text in the image on screen, so you can select it and copy it like text in a document. A photo of a sign, a scanned page or a screenshot becomes text you can paste into a note or a search box.

Recognition runs on your computer. On Windows, pandite uses the text recognizer that is part of the operating system. On macOS and Linux, pandite uses its own recognizer, which is built into the program. Nothing is sent anywhere, and pandite works the same without a network connection.

pandite reads the text after the image is on screen at full resolution and stays there for a quarter of a second. Recognition runs in the background, so the image always comes first. On Windows, it also runs at a low priority. When you move to another image, pandite stops the recognition. pandite keeps the text of recently viewed images until it closes, so an image that you come back to has its text at once.

The recognizer reads upright text best. Sideways or upside-down text gives little or no text. When you rotate or flip the image, pandite reads the text again the way the image now stands. The new text replaces the old text. So a photo that opens on its side gets its text when you turn it upright. The same rules apply: the turned image comes first, stays on screen for a quarter of a second, and then pandite reads it in the background. A crop or a resize does not start a new reading. pandite also keeps the text of each way you turned the image, so when you turn the image back, or undo a turn, the earlier text returns at once.

Select and copy text

1. Open an image that shows text.
2. Move the pointer over a word. When pandite has read the text, the pointer becomes a text cursor.
3. Select the text. Use one of these methods:
   • Drag from the first word to the last word. The selection follows the reading order, from line to line.
   • Double-click a word to select the word.
   • Triple-click a word to select its line.
   • Double-click a word and keep the button down after the second click. Drag to add whole words, from that word to the word at the pointer. Release the button to stop.
   • Triple-click a word and keep the button down after the third click. Drag to add whole lines, from that line to the line at the pointer. Release the button to stop.
   • Press Ctrl+A to select all the text in the image.
4. Press Ctrl+C to copy the selected text. Or right-click and select Copy selected text.

The copied text has one line for each line in the image, and one space between the words of a line.

A drag that starts on the image away from the text pans the image, as before. Click the image away from the text to clear the selection. Esc closes pandite, also while text is selected.

The selection stays on the words when you zoom, pan or crop the image. Words that a crop removes are not selected and not copied. When you rotate or flip the image, the selection stays on the words until the text of the turned image replaces them. Then the selection clears.

Shortcuts

Keys or action	Result
Drag from a word	Select words from that word to the pointer
Double-click a word	Select the word
Triple-click a word	Select the line
Double-click a word, keep the button down and drag	Select whole words from that word to the pointer
Triple-click a word, keep the button down and drag	Select whole lines from that line to the pointer
Ctrl+A	Select all the text in the image
Ctrl+C	Copy the selected text. Without a selection: copy the image
Ctrl+Shift+C	Copy the file path, also while text is selected
Click the image away from the text	Clear the selection
Esc	Close pandite (a menu or dialog closes first)

Platforms and limits

• Windows 10 and 11: pandite uses Windows OCR. Windows reads text in the OCR languages that are installed. Windows installs one with most display languages.
• On Windows without an OCR language, pandite shows no text features and no error.
• macOS and Linux: pandite uses its built-in recognizer, the PP-OCRv6 tiny models of PaddleOCR. It needs no download, no language pack and no network connection.
• The built-in recognizer reads the Latin script, also with accented letters, as in English, German, French, Spanish or Norwegian. It also reads digits, punctuation and Chinese. It knows the plain Greek letters, as in formulas, but not the accented letters of Greek text. It does not read Cyrillic, Arabic, Hebrew or other scripts.
• The built-in recognizer uses one processor core. On a desktop processor, it reads a screenshot of 900 x 420 pixels in about half a second. Large images and pages with many lines take longer, up to a few seconds. The first image after start, and the first image of each new size, take about a tenth of a second more.
• With the built-in recognizer, the box of each word is as high as its line.
• pandite reads still images. It does not read animations.
• Text that is not upright gives little or no text. Rotate or flip the image until the text is upright. pandite then reads it again.
• On Windows, pandite reads large images at 6000 pixels on the long side. On macOS and Linux, it reads them at 2000 pixels. Very small print in a very large photo can be lost. Zooming in does not change this.
• The models of the built-in recognizer are licensed under the Apache License 2.0. See THIRD-PARTY-NOTICES.md.

Add an OCR language on Windows

1. Open Settings. Select Time and language, then Language and region.
2. Select Add a language, then select the language.
3. In the optional features, make sure that Optical character recognition is selected.
4. Select Install.
5. Restart pandite.

pandite reads the languages of your user profile that have OCR data.

Turn text recognition off

1. Close pandite.
2. Open settings.ini in a text editor. The README gives its folder for each system.
3. Find the line ocr = auto and change it to ocr = off. If the line is missing, add ocr = off at the end.
4. Save the file and start pandite.

With ocr = off, pandite reads no text and shows no text features.

This site is open source. Improve this page.