Prepare a readable image
Use a sharp image with readable contrast. Select English, Simplified Chinese or both. For a large screenshot, crop to the relevant area or resize it in Image Toolbox before recognition.
Use the recognized text
A typical workflow is a cropped screenshot of an English error message: upload the PNG, select English, click Extract text, and compare every identifier and number with the image. For mixed English and Simplified Chinese, select both languages. Recognition can confuse 0 with O or omit punctuation even when most of a line looks correct.
| Source | Supported workflow |
|---|---|
| PNG, JPEG or WebP screenshot | Upload the image and select its language |
| PDF document | Export the relevant page to an image first; direct PDF input is unsupported |
| Rotated, blurry or very small text | Improve the source crop and review recognition manually |
Copy or download the text, count words, or pass it to the Sensitive Data Redactor for review. OCR does not preserve a document layout or promise to reconstruct source code exactly. Always check punctuation, identifiers and numbers.
Resources and limits
For offline recognition, save the basic tools and the text recognition pack before disconnecting. The pack includes both supported languages.
Recognition loads local language models on demand and may take longer on a phone. Images are limited to ten megabytes and twelve megapixels. The job can be cancelled and has a two-minute timeout. No image, text or automatic OCR history is saved.
Use Image Toolbox to resize a screenshot, Remove Empty Lines to clean the reviewed text, or Sensitive Data Redactor before sharing it. To extract addresses from reviewed output, use Email Extractor. Redaction and extraction also require review; OCR mistakes can hide a sensitive value from pattern matching.
Powered by open-source software. Third-party licenses.
Documentation examples checked: . How these tools are checked.
Frequently asked questions
Is the image uploaded to an OCR provider?
No. The worker, WebAssembly engine and language models load from this site. Recognition happens locally and images are not sent to an OCR service.
Can I recognize text in a PDF?
This tool accepts PNG, JPEG and WebP images. It does not directly accept PDF documents. Export or capture the relevant page as an image first.
Is the recognized text guaranteed accurate?
No. Small text, compression artifacts, unusual fonts, complex layouts and code punctuation can cause recognition mistakes. Confidence is an engine score, not a guarantee. Review the text before using it.
Sources
Spotted a mistake? Email hello@devtoolsmith.com and we’ll fix it.