Image-to-text tool
Read printed text from an image with Tesseract.js. Everything happens in your browser.
Ready on your device
Choose an image you have the right to use. It is not uploaded.
Start here
Extract printed English, Spanish, French, or Simplified Chinese text.
Drop an image here
or
PNG, JPG, WebP, or BMP up to 25 MB. Clear, level text gives the best result.
Selected file
Choose the main printed language. OCR reads text; it does not translate it.
Default is recommended for everyday images and faster startup.
Best + Preprocess workflow
Tesseract.js 5.0.5 runs locally
First use downloads the selected language model to your browser. Your image is not sent with that request.
The OCR engine and language data are loading in this browser.
Finished on your device
Review the result before copying or saving it.
Image decoding and OCR happen in this browser tab. Only the OCR runtime and selected language data are downloaded.
A clear local workflow
Choose an image, identify its printed language, and select the processing mode that fits the image and your device.
Select a PNG, JPG, WebP, or BMP that you own or may process. Choosing it gives only this browser tab access to the file.
Select English, Spanish, French, or Simplified Chinese. Use Default for speed, Best for higher-accuracy language data, or Best + Preprocess for automatic enlargement, grayscale, contrast enhancement, OCR, and conservative spelling cleanup.
Correct names, numbers, punctuation, and reading order, then copy the result or save a plain TXT file.
Default starts fastest for clear screenshots. Best uses higher-accuracy language data without changing the image. Best + Preprocess uses the Best data after locally enlarging small images, converting them to grayscale, and stretching contrast, then performs conservative text cleanup.
Tesseract.js and its WebAssembly core are served from NoUpload App. The chosen language file comes from the dedicated NoUpload App model host. Your image is read from browser memory and is not uploaded with those requests.
Default works best with a current 64-bit browser, a dual-core CPU, and at least 4 GB RAM. For Best, use 8 GB RAM and four logical CPU cores. For Best + Preprocess, 8 GB is the practical minimum, while 16 GB RAM and four or more modern CPU cores are recommended. A dedicated GPU is not required.
For Best + Preprocess, keep images near 12 megapixels or below for smoother work. Very large photos can use hundreds of megabytes of temporary browser memory. Phones and low-memory tablets should use Default with a cropped image. Keep the tab open until processing finishes.
The final cleanup fixes a small set of characteristic OCR substitutions, broken word joins, spacing, and punctuation. For Chinese, it focuses on incorrect spaces between Han characters and nearby punctuation. It is not a full dictionary, grammar checker, translator, or factual verifier, so always review names, codes, prices, and dates.
Tesseract is primarily for printed text. Handwriting, curved pages, glare, shadows, perspective distortion, decorative fonts, tiny characters, multi-column reading order, tables, formulas, and mixed-language pages can still produce errors. Automatic enhancement cannot recreate detail that is missing from the original image.
Straight answers
Useful details before reading text from an image.
No. The browser reads your image locally and sends it only to the Tesseract worker in this tab.
OCR needs trained character and language data. The selected gzip file downloads toward your browser; your image is not included in that request.
Default prioritizes speed. Best uses higher-accuracy trained data on the original image. Best + Preprocess uses the same Best data after automatic grayscale, size, and contrast preparation, then applies conservative post-OCR text cleanup. The Best language download is shared by the last two modes when browser caching is available.
It creates a temporary OCR-only canvas in browser memory, enlarges small images when useful, converts pixels to grayscale, measures the brightness range, and automatically improves contrast. It does not overwrite or upload the original file.
No. It is deliberately conservative and fixes only recognizable OCR artifacts, spacing, and punctuation patterns. It does not understand every word or sentence, and Chinese cleanup does not attempt broad dictionary replacement. Review the result before using it.
Use a current 64-bit Chrome, Edge, Firefox, or Safari. Default is suitable for a dual-core computer with 4 GB RAM. Best is more comfortable with 8 GB RAM and four logical cores. Best + Preprocess is best with 16 GB RAM and four or more modern cores, especially for large images.
English, Spanish, French, and Simplified Chinese are available. Choose the language that best matches the printed text.
Tesseract is designed primarily for printed text. Handwriting results are generally unreliable.
The page uses Tesseract.js 5.0.5 with its WebAssembly OCR core. Tesseract.js and the bundled core are provided under the Apache License 2.0; license copies are available from the site’s Open-source Licenses page.
Yes. Screenshots and clear scans work well when text is large enough and the image is not blurred or heavily compressed.
No account or sign-in is required, and the tool is free to use.
More local tools
