Image OCR

🔍 Image OCR Text Recognition

This tool pulls text out of images. Upload a picture containing text and the local OCR engine reads it, returning the lines as editable text. Simplified Chinese, English, and mixed Chinese-English content are all supported. The recognition engine and its language models ship inside the plugin, so your images never travel to a server.

Before recognition you can choose a preprocessing mode: keep the original, convert to grayscale, or convert to black and white. Screenshots with busy backgrounds and heavy noise usually read better after binarisation, while photographs often do better untouched. When you upload several images at once, each is processed in turn and the results are combined.

It suits scanned documents, screenshots, and photos of paper that need to become editable text; grabbing a phone number or address embedded in an image; digitising a paper form; or copying text out of a picture where selection is impossible.

Because everything runs locally, the first recognition has to load the engine and the model, which takes a moment. Subsequent runs in the same page reuse what is already loaded and feel considerably faster. Nothing is cached between visits, so the wait returns on a fresh page load.

Results are returned per image rather than as one merged block, so you can tell which line came from which file when working through a batch. Very wide images are downscaled before recognition, which keeps memory use sane but can cost a little accuracy on small type.

Rotated pages are not auto-corrected, so if the text comes out scrambled or empty, rotate the image first and try again.

Rotated or upside-down pages are not auto-corrected, so when the output looks scrambled or comes back empty, rotate the image in any viewer first and run it once more.

✨ Features

• Reads Simplified Chinese, English, and mixed Chinese-English text
• Engine and language data are bundled locally: no upload, works offline
• Three preprocessing modes: original, grayscale, black and white
• Batch upload, processing several images in sequence and merging the output
• Copy the recognized text with one click, or download it as txt
• Progress is shown, so large images and batches never look frozen

🚀 How to Use

1. Click to choose an image, or drag it into the upload area
2. Select the language: English, Simplified Chinese, or both
3. Pick a preprocessing mode; black and white suits screenshots
4. Press start and wait for the progress to finish
5. Review the result, fix any errors, then copy or download it

⚠️ Notes

• First load takes a while: the engine and language data are sizeable, so the initial run pauses to load them; later runs in the same page are much quicker
• Clarity drives accuracy: blurry, skewed, or low-resolution images degrade results noticeably, so start from the sharpest original you have
• Handwriting is out of scope: the engine is trained on printed text, so handwriting, decorative typefaces, and unusual fonts read poorly
• Layout is not preserved: output is plain text arranged by line, so table structure, columns, and image placement are lost
• Always proofread: misreads happen, and anything involving amounts, ID numbers, or other critical values deserves a character-by-character check