Image to text
Choose a screenshot, recognize locally, then review and edit.
PNG, JPEG or WebP · up to 10 MB and 24 megapixels. Large images are reduced to a bounded preview. Crop tightly for small lettering.
Your image and text stay in this browser. No upload or account.
First recognition downloads about 6.6 MB of engine and language files from this site. Recognition resources load only when you start. Later runs can use your browser’s resource cache.
0 characters
Coverage & counts
0 characters · 0 UTF-16 units
From screenshot to editable text, in three steps
Choose a PNG, JPEG or WebP, paste a screenshot into the upload area, or drop one there. Select English for Latin text; select Simplified Chinese + English for a mixed screenshot. Rotate sideways text before starting. Crop the image in your screenshot or photo editor to remove unrelated text and improve small lettering.
The preview uses a reduced image when the source exceeds the processing bounds. Clear image & text removes the current image and result from the tool. Cancel recognition stops the worker, including during engine loading, so you can change the image and try again.
Review the words before using them
OCR can confuse 0 with O, 1 with l, punctuation and names. The confidence indicator is the engine’s estimate, not a checked accuracy percentage. Edit the result above before copying a number, web address or quotation. Blank images may produce no text; handwritten notes, textured backgrounds and decorative fonts can produce mistakes even when a screenshot looks readable to you.
This tool is intended for printed text and clear screenshots. It does not understand a document’s meaning or promise to reconstruct columns, tables or original formatting. PDF, SVG and animated formats are not supported. For a PDF page, take a screenshot of the portion you need first.
Your image stays here; the recognition engine loads on demand
The image is decoded and recognized in this browser with a local WebAssembly engine. Your image and recognized text are not uploaded or saved as drafts by this tool. The first run downloads the engine and the selected language files from this site. Starting again can use normal browser caching; those downloads contain the engine and dictionaries, not your image.
A slow connection or limited device can make the first run take longer. Use Cancel recognition if you want to stop. If resources cannot load, retry after checking your connection. After the necessary files have loaded, recognition itself does not need a remote OCR service.
Give the extracted text its next job
Use the tool links beside the result to take your edited text into fancy alphabets, bold emphasis or a static QR code. The handoff stays in memory during navigation; it is not placed in the page URL. Image-title tools accept shorter text and show their limit.
If you already have copied styled characters rather than a screenshot, the fancy-to-plain converter can restore supported alphabets without OCR. For connected handwriting as an image, use the handwriting image generator.
Frequently asked questions
Which languages can I recognize?
Choose English or Simplified Chinese + English. Other languages are not offered in this version. Selecting the wrong language can reduce the quality of the result.
Can it read cursive handwriting?
It may pick out some clear letters, but connected handwriting and decorative typefaces are unreliable. This tool is designed for printed text and screenshots; always review what it returns.
What are the image limits?
PNG, JPEG and WebP files must be at most 10 MB and 24 megapixels. Processing is reduced to at most a 3,000-pixel edge and about five megapixels to limit browser memory use. A tight crop can preserve the lettering better than a huge full-page image.