ImgBox

Image to Text (OCR)

Online image-to-text (OCR) tool: extracts Chinese/English text (and emoji) from screenshots, photos, or scans into editable text you can copy or download as TXT. First load downloads ~30 MB of model + runtime; subsequent runs are instant and work offline. Images are never uploaded.

Photo skewed or cluttered around the edges? Deskew with Document Scan first →

OCR is not supported on this device

The OCR engine needs more memory than iOS Safari and in-app browsers (WeChat, QQ, Douyin, Xiaohongshu, etc.) allow per tab, so it can't run here. Please open this page on a desktop browser (Chrome, Edge, Firefox, or Safari on macOS).

Related Tools

🔍QR Code ReaderTAdd Text to Image◐Compress Image

How to use

Purpose

Image-to-text OCR running entirely in your browser via Tesseract.js — recognizes Chinese, English, Japanese, Korean, and mixed-language text from photos, screenshots, and documents. Outputs editable / copyable text. The OCR model runs on-device, so contracts, notes, business cards, IDs, and sensitive documents are never uploaded. Supports Simplified / Traditional Chinese.

Steps

  1. Upload an image containing text
  2. Pick language: Simplified Chinese / Traditional Chinese / English / Japanese / Korean / mixed
  3. First use: download the language model (Chinese ~14MB, English ~4MB)
  4. OCR engine runs in-browser (1-10 seconds depending on image size)
  5. Preview the extracted text
  6. Copy to clipboard / download as TXT

FAQ

How accurate is the OCR?
Printed text in clear, large fonts: 95%+ across Chinese, English, Japanese, Korean. Handwriting: 60-80% (much lower). Skewed / blurry / poorly-lit photos drop accuracy — preprocess with doc-scan first.
Are my images uploaded?
No — Tesseract.js and language data run in your browser. Critical for business cards, contracts, sensitive notes.
Languages supported?
Currently: Simplified Chinese, Traditional Chinese, English, Japanese, Korean. Mixed-language (e.g. Chinese + English) also works.
Tables and formulas?
Tesseract excels at plain text. Simple tables get recognized but cell structure isn't preserved; math / chemistry formulas have low accuracy. For those, use Mathpix or specialized tools.
vs Google Cloud Vision / Baidu OCR?
Cloud OCR (Google / Baidu / Tencent) uses larger models with higher accuracy, but requires upload and has quotas. Our advantage: fully local + no quota + privacy. Trade-off: slightly lower accuracy, but privacy is the default.

Use cases

  • Business card → contact info extraction
  • Contract / agreement photo → editable text
  • Notes / whiteboard → searchable text
  • Book / article scan OCR for editing
  • ID number / receipt info extraction

Use cases

Chinese / English / Japanese / Korean printed-text OCR — model on-device, image never uploaded. Business cards / contracts / notes — privacy by default.