Image to Text Converter (OCR)
Extract text from photos, scans, and screenshots using real optical character recognition. Supports seventeen languages, multi language documents, and automatic image cleanup for better accuracy.
Click or drop an image (JPG, PNG, WEBP, BMP)
Select one or more languages present in the image. Selecting extra unused languages can slow recognition down slightly.
A production OCR engine, compiled to WebAssembly, running in your tab
This tool is a thin interface over Tesseract.js, the JavaScript and WebAssembly port of Google’s Tesseract OCR engine, the same open-source engine used in countless document scanning pipelines. The library is loaded once from a CDN script tag, and from that point on, every image you feed it is recognized locally, inside the browser’s WASM sandbox. No image, and no extracted text, is sent to a server. It’s a genuinely heavy piece of software to run client-side, which is exactly why the tool shows a real progress bar instead of a spinner, recognition on a full page of text can take several seconds even on a fast machine.
What happens between upload and text
Tesseract.recognize(), which loads the matching trained language data and runs its own internal pipeline: binarization, layout analysis, line and word segmentation, then character recognition.
Stretching contrast this way is a simple normalization: whatever the actual darkest and lightest grays in the photo are, they get remapped to true black and true white, and everything in between is scaled proportionally. On a scan with a dull gray background instead of pure white, this alone can noticeably improve character edge detection before Tesseract even starts.
Language support
| Feature | Detail |
|---|---|
| Languages offered | 17, selectable independently as checkboxes |
| Multiple languages | Joined with + and passed as one string, e.g. eng+fra |
| Default | English if nothing is checked |
| Model download | Per-language trained data, fetched once and cached by the browser |
Confidence score
A single number from Tesseract’s internal per-character confidence estimates, averaged. Low scores usually mean low resolution, poor contrast, or an unusual font, not a bug in the tool.
Preprocessing toggle
Off by default because clean, high-contrast scans don’t benefit from it and it adds a canvas round trip. Worth enabling for phone photos of receipts or low-contrast printed pages.
Text output actions
Copy to clipboard uses the standard Clipboard API; download saves the raw extracted text as a .txt file built from a Blob, no server round trip for either.
- Tesseract.js, the pure JavaScript/WASM port of Tesseract this tool is built on.
- Tesseract OCR, the original open-source engine originally developed at HP and now maintained under Google’s sponsorship.
- Tesseract trained data files, the per-language models Tesseract.js downloads on demand.
Text people pull out of images
Pulling text out of a scanned receipt for expense tracking, digitizing a printed page from a book or document you don’t want to retype, extracting a quote from a screenshot, converting a photographed whiteboard into editable notes, or lifting text from a foreign language sign or menu photo using the right language pack. Because it’s fully client side, it also works for sensitive documents you’d rather not upload to a cloud OCR service.
FAQ: Image to Text Converter (OCR)
Accuracy depends heavily on image quality. Clear, well lit, high resolution text typically recognizes at ninety five percent or higher. Blurry photos, low contrast scans, or handwriting reduce accuracy, handwriting recognition in particular is not this engine’s strength. The result screen shows a confidence score so you can judge each extraction.
Yes, check every language present in your image before extracting. The engine loads a trained model for each selected language and combines its analysis, which works well for bilingual documents but does add a small amount of extra processing time per additional language.
It converts your image to grayscale and stretches the contrast so dark text stands out more clearly against its background. This is especially helpful for phone photos with uneven lighting or shadows, and can noticeably increase the confidence score compared to running the raw image directly.
Yes, the result appears in an editable text area, so you can fix any misread characters before copying it to your clipboard or downloading it as a plain .txt file.
No. Recognition runs through a WebAssembly build of the Tesseract OCR engine loaded into your browser. Your image is processed on your own device and never transmitted anywhere.
From the blog
Notes on images and colour
Format choices, compression tradeoffs and accessible palettes.
