🌍 Multi-language & Handwritten Text Extraction Demo

Powered by DeepSeek-OCR | Extract text from images and PDFs in multiple languages

✨ Support for handwritten text, tables, and 12+ languages with confidence scoring

📤 Input Configuration

⚙️ Processing Settings
Language
Processing Mode
Task Type

📊 Results

ℹ️ Information

🔧 Processing Modes

  • Gundam: 1024 base + 640 tiles with cropping - Best balance
  • Tiny: 512×512, no crop - Fastest
  • Small: 640×640, no crop - Quick
  • Base: 1024×1024, no crop - Standard
  • Large: 1280×1280, no crop - Highest quality

📋 Task Types

  • Markdown: Convert document to structured markdown (grounding ✅)
  • Free OCR: Simple text extraction
  • Locate: Find specific text in image (grounding ✅)
  • Describe: General image description
  • Handwritten: Specialized handwritten text extraction (grounding ✅)
  • Table Extract: Extract and format table data (grounding ✅)
  • Custom: Your own prompt (add <|grounding|> for boxes)

🌍 Language Support

Supports 12+ languages including English, Arabic, Urdu, Chinese, Japanese, Korean, Spanish, French, German, Hindi, and Russian with automatic language detection.

💡 Tips

  • Use Gundam mode for best results
  • Handwritten task works best for handwritten documents
  • Table Extract automatically formats tables into markdown
  • Confidence scores show model certainty for each detected region
  • PDF processing supports up to 50 pages