Skip to content
Kitbay
Skill

paddleocr-text-recognition

Apache-2.0Updated on Sep 16, 202692,573GitHub starsNeeds an API key

The PaddleOCR project's skill for pulling plain text out of images, photos, scans, screenshots and scanned PDFs.

What the skill teaches

The PaddleOCR project's skill for pulling plain text out of images, photos, scans, screenshots and scanned PDFs. The agent runs the paddleocr command-line tool's api subcommand on a file path or URL, which sends the file to PaddleOCR's official API, optionally picking a model such as PP-OCRv5, page ranges and whether to correct warped or rotated pages. It returns line-level text with confidence scores per page as JSON. It needs a PADDLEOCR_ACCESS_TOKEN, reports authentication, rate-limit and no-text errors directly instead of falling back to the agent's own vision, and points to the separate document-parsing skill for tables, formulas or complex layouts.

Capabilities

  • Extract text lines from images and scanned PDFs
  • Return recognition scores for each line
  • Process selected PDF page ranges