Image to Text (English OCR)
Extract editable English text from a still image, privately in your browser.
What Image to Text (English OCR) helps you do
Copy the printed English words in a screenshot, scanned image or clear photograph without retyping them. This tool runs the open source Tesseract OCR engine on your device and produces editable plain text.
How to use it
- Choose one still JPG, PNG or WebP image, or drop it into the input area.
- Select automatic layout, one block or one line. Optionally enable grayscale preprocessing for colored images.
- Select Extract text. The English engine and model load from this site on demand, then recognition runs in a background worker.
- Compare the editable result with the source image, correct any errors, then copy it or download a UTF-8 TXT file.
Input and output example
A clear screenshot reading “Meeting agenda” on the first line and “Budget review at 10:30” on the next.OUTPUT
Meeting agenda Budget review at 10:30
Supported formats and limits
- One still JPG, PNG or WebP, up to 5 MiB (5,242,880 bytes), 8 megapixels and 8,000 pixels on either side. Animated images and PDF files are not accepted.
- Only the English recognition model is included. Handwriting, small or blurred characters, decorative fonts, unusual reading order and other languages may be inaccurate.
- The first run downloads several megabytes of engine and English model files from this site. It needs a modern browser with WebAssembly and Worker support; slower phones may take longer or run out of memory.
- Plain text is returned. Font styling, columns, tables and exact spacing are not preserved. Review names, numbers and punctuation before relying on the result.
How the result is calculated
The browser applies embedded image orientation, then the image is drawn on a white background at its original pixel size. Optional grayscale changes only the recognition copy. Tesseract uses its LSTM English model with automatic, single-block or single-line page segmentation. The displayed confidence is the engine’s internal 0–100 estimate, not a measured accuracy percentage or probability that the text is correct. Inputs and results stay in memory until reset or leaving the page; no OCR data is uploaded or saved by this tool. Public engine/model assets may remain in the browser’s ordinary HTTP cache. The engine’s IndexedDB model cache is disabled.
Frequently asked questions
Is this OCR free, and do I need an API key?
The tool needs no account, API key or paid AI service. Recognition runs on your device using Tesseract.js. Your normal internet and device costs still apply.
Does my image leave my device?
No. The image and recognized text are processed locally. Your browser requests only this site’s public app, engine and English model files. Reset or leave the page to release the tool’s working data.
Why did OCR miss a word or read a number incorrectly?
OCR guesses characters from pixels. Crop empty margins, straighten the page and use a sharp, well-lit image with larger text. Try single block or single line when automatic layout misses short text. Always proofread the result.
Can I extract Korean text or make a scanned PDF searchable?
This version includes only English image OCR. It does not offer PDF OCR or create searchable PDFs. For a PDF page, export a still image with the PDF to Images tool first; other languages need additional compatible models.
Can I stop a long recognition task?
Yes. Cancel terminates the OCR worker, including engine initialization. Your selected source remains available for another attempt. Start over also clears the image and result.