Online OCR

Automatic Chinese, English and Japanese OCR with local GPU acceleration. Recognize long images or regions, review text boxes, edit and export TXT. Images stay on your device.

Upload · Adjust · Export
Automatic multilingual OCR

Simplified / Traditional Chinese, English and Japanese. No language selection.

Recognition area

Upscale small images for detection; split long images at source resolution.

Images stay on this device. First use loads about 58 MB of engine and models, then uses browser caching.

For screenshots, scans and mildly tilted text. Review handwriting, severe blur, table order and complex columns manually.

Recognized text

Online OCR · What this tool does

PP-OCRv5 detects and reads text locally in your browser. Simplified and Traditional Chinese, English and Japanese can be mixed without language selection. Upload or paste, recognize the whole image or a region, inspect text boxes and confidence, then edit, copy or export TXT and boxes JSON. Long images use overlapping source-resolution tiles. Automatic mode prefers local WebGPU and retries on CPU when unavailable or unsuccessful.

How to use it

  1. Upload PNG, JPG or WebP, paste a screenshot or try the multilingual sample.
  2. Use automatic multilingual recognition; select a region, rotate or enhance small text if needed.
  3. Recognize, review low-confidence lines against the image and edit or export the full text.

Practical tips

  • Limits: 25 MB, 12 megapixels and 16000 px per edge. Long screenshots use overlapping tiles with duplicate boxes removed.
  • First use loads about 58 MB of static engine and models. Images are not uploaded. You can force CPU or WebGPU instead of automatic mode.
  • Use the scrollable 100% preview for small text. Review tables, columns, handwriting and severe blur. The model does not cover every language.

Frequently asked questions

Do I need to select a language?

No. A joint model reads Simplified and Traditional Chinese, English and Japanese, including mixed images. This is not a universal recognizer for all languages.

Can I use OCR without a GPU?

Yes. Automatic mode uses local CPU when WebGPU is unavailable or fails. Current Chrome or Edge can use WebGPU; CPU processing may take longer.

Creative workspace / ImgHubExplore all tools