OCR image to text
Pull text out of screenshots, scans and photos in Chinese, English, Japanese, Korean, German or French, without uploading them.
Or choose them. Files are processed in this tab and never uploaded to any server.
Copying text out of a screenshot, a scanned contract or a photo of a slide is the kind of job that should not require sending the image to someone else's server. This tool runs Tesseract in the tab through naptha/tesseract.js, the WebAssembly port of the engine: choose the language, drop one image or several, and get the text with the engine's own confidence score. Simplified Chinese plus English is the default, since Chinese screenshots so often contain English terms.
How it works
- The first run downloads the tesseract.js worker, the WASM core and the language model from the jsDelivr CDN; the browser caches them, and the image itself never leaves the tab.
- Several images are recognised one after another with a single worker, and each result is labelled with its file name.
- Tesseract puts spaces between Chinese and Japanese characters; they are removed where both neighbours are CJK, so mixed text such as GPU 显存 keeps its gap.
Where your data goes
Nowhere. This tool runs entirely in your browser: the text you paste is processed by the page and is never transmitted to a server or written to a log.
This tool is free and needs no account. Its results exist only in your open page and are not saved anywhere.
What it costs
This tool is free, with no sign-in and no points.
Common questions
- How can I get better results?
- Tesseract does best on clean, printed, horizontal text with the characters at least about 20 pixels tall. Crop to the text, use a screenshot rather than a photo of a screen, and pick the exact language — Traditional Chinese recognised as Simplified produces substitutions. Handwriting and heavily stylised fonts are beyond what this engine does well.
- What does the confidence number mean?
- It is Tesseract's mean word confidence for the page, from 0 to 100. Above about 90 the text is usually right; in the 60s and 70s expect wrong characters, especially similar-looking Chinese ones. It measures the engine's certainty, not accuracy, so always proofread anything that matters.
- Why is the first run slow?
- The engine and language data are downloaded on first use, several megabytes for Chinese, and cached afterwards. If jsDelivr is slow or blocked on your network, the download stalls and the tool cannot run; after one successful run it works again from the cache.
The open-source behind it
This tool runs on naptha/tesseract.js, released under Apache-2.0. If you need the same behaviour inside your own program, that is the library to reach for.
naptha/tesseract.jsAlso known as
- image to text
- ocr online
- extract text from image
- screenshot to text
- chinese ocr
- tesseract online