Try Pro FREE for 2 weeksGet Started
Start for free

PaddleOCR — OCR AI Model

PaddleOCR reads printed text and needs no GPU at all, which makes it the one to run over a large pile of documents. The lang parameter picks the recognizer: Korean (Hangul + alphanumerics), Japanese (Japanese + alphanumerics), Chinese (Chinese + alphanumerics) or English (English letters + digits). Printed text only - for layout, tables or handwriting, use GLM-OCR or DeepSeek OCR instead.

Try in Studio

What it does

PaddleOCR reads printed text out of an image and hands it back as text you can copy and search - a menu, a sign, a screenshot, a page of a document.

Set lang to match the script you are reading: Korean for Hangul, Japanese for Japanese, Chinese for Chinese, English for English. Every value also reads alphanumerics; English reads Latin text only. Japanese and Chinese share one multilingual recognizer.

It needs no GPU, which is what makes it the one to point at a large pile of documents. It reads printed text only - for layout, tables or handwriting, GLM-OCR or DeepSeek OCR is the better call.

When to use it

  • A pile of printed documents worked through in bulk, on CPU
  • Menus, street signs and advertisements read off a photograph
  • Text lifted out of a screenshot instead of retyped by hand
  • Printed Korean, Japanese, Chinese or English text, picked with lang
  • Pulling text out to hand to a translation step afterwards

Details

  • CategoryOCR
  • Year2025
  • LicenseApache-2.0

Compliance & Provenance

  • ProviderOpen-source · Specialized
  • EU AI Act RiskMinimal Risk
  • Art. 50 TransparencyNot applicable

Inputs & Outputs

ImageInputimage

Source image containing text to recognize

TextOutputstring

Detected texts from an image

Region DataOutputstring

JSON: one record per detected line with its text, recognition score, 4-point polygon, and bounding box

Parameters

langstringdefault Koreanoptions: Korean, Japanese, Chinese, English

Recognition language. Korean = Hangul + alphanumerics, Japanese = Japanese + alphanumerics, Chinese = Chinese + alphanumerics, English = English letters + digits. Printed text only.

Recognition language for this OCR node. Printed text only for every value - handwriting is not supported; use GLM-OCR or DeepSeek OCR for handwritten input. "Korean" (default, 한국어 / 한글): Hangul plus alphanumerics, the Korean-tuned recognizer. "Japanese" (일본어): kanji and kana plus alphanumerics. "Chinese" (중국어): Chinese plus alphanumerics. Japanese and Chinese share PaddleOCR's multilingual server recognizer, so they behave identically. "English" (영어): Latin letters and digits only - it will not read Hangul, Japanese or Chinese. For Japanese documents with vertical text, tables or formulas use Sarashina2.2-OCR instead.

Tags

  • ocr
  • text-recognition
  • multilingual
  • korean
  • japanese
  • chinese
  • english
  • text-extraction
  • printed-text
  • no-gpu
  • fast

Alternatives in Text Recognition(OCR)

Resources

All AI models