Windows workflows
OCR — Image to Text
Multi-language recognition, browser-based
🔒 See this tool’s privacy notes for local processing and external service details. Site analytics are described in our privacy policy.
🔍
Upload image
ℹ️ First use downloads a language model (3-15MB). Subsequent uses are cached and fast. Everything runs in your browser.
Quick guide
What can OCR (Image to Text) help you do?
Free online OCR. Extract text from photos, screenshots, scanned documents. Supports Chinese (Traditional/Simplified), English, Japanese, Korean. tesseract.js browser-based, no upload.
Best for
Use this to extract copyable text from a clear screenshot, printed page, or scanned image. Traditional Chinese, Simplified Chinese, English, Japanese, Korean, and Chinese-English combinations are available. It is not a document converter that preserves table structure or original page layout.
How to use it
Choose or drop one image, select the language matching its contents, and start recognition. The default follows the page language; use a combined option for Chinese-English text. Once complete, copy or download TXT. Clear removes the current result and selected-image state.
How to read the result
The output is plain text returned by Tesseract, potentially with incorrect characters, extra line breaks, or changed reading order. Progress measures work completed, not recognition accuracy. The result field is read-only, so copy it into another editor for corrections.
Privacy and limits
The local file is passed to a browser Tesseract worker rather than an upload-based recognition API. The script loads from jsDelivr, and worker, core, and language-model resources may require additional downloads. Model caching is managed by the dependency and browser.
Handwriting, blur, skew, tiny type, and multiple columns can produce errors; compare important numbers with the source. Changing images or languages, or clearing, removes old results and prevents outdated work from restoring text. Wait for successful recognition of the current image before copying or downloading.
Frequently Asked Questions
How accurate?
Accuracy depends on image quality, language, layout, and font. Verify important output; handwriting and blurry images are harder.
Why is the first run slow?
Downloads a language model (Chinese ~12MB, English ~3MB). Cached after that.
Does it support Chinese?
Yes, both Traditional and Simplified. The Chinese model is larger (~12MB).
Is my image uploaded?
No. Uses tesseract.js (WebAssembly in your browser).