Skip to main content
OS Authority

Image to text converter for printed English

Recognize printed text in a screenshot, photo or scan on your device. Compare the source preview with the editable transcript before copying or downloading it.

Your images and recognized text are processed on this device. Their contents are never uploaded or included in analytics.

How it works

  1. Choose clear JPG, PNG, WebP or HEIC/HEIF images, or use the camera or image clipboard where available.
  2. Rotate sideways text if needed, then choose Extract text. The first recognition loads the English OCR model.
  3. Compare the source previews with the transcript and correct names, numbers and punctuation.
  4. Copy the edited transcript or download it as a UTF-8 TXT file.

Prepare a clear source

Printed text works best when upright, sharp and evenly lit. A screenshot usually gives better results than a photo of a screen. Crop away unrelated areas with Crop Image before importing if needed. For a letter saying Invoice 1024, check that the transcript has not mistaken 1 for I or 0 for O. Blurry text, unusual fonts, handwriting and complex tables can fail.

Review every important detail

OCR is an estimate of what the image says, not an exact copy. The confidence value is the recognizer's 0–100 internal score; it is not a measured accuracy percentage or a guarantee. A high score can still accompany a wrong name or amount. Read the source alongside the transcript and confirm critical details before using them.

Recognition happens locally

An English model and processing code are downloaded from this site's own origin when recognition starts. Image contents stay in browser workers. Batch images are recognized one at a time to bound decoded-image work. Device limits reject excessive files, dimensions, cumulative pixels or transcripts instead of promising unlimited processing. Cancel terminates the active recognizer and its processing worker.

Frequently asked questions

Can I extract text from a screenshot?

Yes. Import a JPG, PNG, WebP or HEIC/HEIF image or paste an image from the clipboard where your browser permits it. Text must be printed English for this version.

Does it recognize handwriting or reconstruct tables?

That is not supported. Some characters may be recognized, but handwriting is unreliable and a plain-text transcript does not recreate table columns or document formatting.

Why did no text appear?

The image may contain no recognizable printed text, or the text may be too small, blurry or sideways. Try a sharper crop with larger upright letters. A blank result is not proof that the image contains no text.

Will this work offline?

The processing code and model must be available first. If those assets cannot load, recognition will fail with an offline or loading error. We do not promise that every browser retains the large model between visits.

Last reviewed: October 11, 2026

Related tools