Image to Text

How to use

1. Drop or paste one JPG, PNG, WebP, or BMP image.
2. Choose English, Korean, or Korean + English to match the visible text.
3. Wait for the on-device OCR progress to finish, then review the editable result.
4. Copy the text or download it as a UTF-8 TXT file.

What happens when you extract text from an image

OCR, short for optical character recognition, turns pixel shapes into selectable characters. This tool runs the Tesseract LSTM OCR engine as WebAssembly in the browser. The image remains in the current tab while the engine reads it; the worker, engine, and English and Korean language files are served from this site rather than sending the image to an OCR API. The selected language model is loaded only when you use it, and the browser can cache it for later scans.

The result is plain editable text, not a reconstruction of the picture. Paragraph breaks are kept where the engine finds them, but columns, tables, font styling, and the exact position of labels do not survive in TXT. Read through names, dates, totals, and similar-looking characters such as I, l, and 1 before using the result in a form or record.

Choose the language that is actually in the image

Use English for English-only screenshots, URLs, source code, and printed English documents. Use Korean for Korean-only pages. For a Korean document containing English product names, figures, or web addresses, choose Korean + English so both character sets are available to the recognizer. The mixed setting loads both models, so its first use takes longer and it may be less decisive than a single-language model when the image clearly contains only one language.

This tool accepts one JPG, PNG, WebP, or BMP image up to 20 MB and 24 million pixels. Those limits keep the browser responsive on phones and laptops; they are not limits of the image formats themselves. A 12-megapixel phone photo is within the pixel limit. For HEIC photos, convert the file to JPG first with the related HEIC to JPG tool.

Get a cleaner OCR result from photos and screenshots

OCR works best on sharp, high-contrast, straight-on printed text. A cropped screenshot often reads more cleanly than a full desktop image because navigation bars, icons, and decorative backgrounds are removed. For a phone photo, fill the frame with the page, avoid glare and shadows, and keep the camera parallel to the text. Large text and one-column layouts generally produce more reliable reading order than small multi-column pages or dense tables.

Handwriting, cursive, low-resolution images, stylized fonts, reflections, and tilted pages need careful checking. The confidence value is an engine-wide estimate, not proof that every word is correct. Edit the textarea before copying; the copied and downloaded versions contain your edits. If the source is an existing digital PDF with selectable text, use PDF to Text instead, which reads its text layer directly without OCR.

Frequently asked questions