What should I use for OCR on invoices if I need line items and totals?

For invoice OCR with line items and totals, you’ll usually want more than “plain OCR.” You need document understanding / invoice extraction that can read the page and return structured fields.

Best options

• Google Document AI – Invoice Parser

  • Strong for invoices, line items, totals, taxes, vendors, dates
  • Good accuracy and structured output

• Azure AI Document Intelligence (Form Recognizer) – Prebuilt Invoice

  • Good for extracting invoice fields and line items
  • Easy to integrate if you’re already in Microsoft/Azure

• Amazon Textract – AnalyzeExpense

  • Specifically designed for invoices/receipts
  • Extracts line items, totals, and key values

• ABBYY Vantage / FlexiCapture

  • Often excellent for complex enterprise invoice workflows
  • Strong if you need customization and human review steps

If you want open-source / self-hosted

• PaddleOCR + custom extraction logic
• Tesseract + layout parsing, though this is usually weaker for line items
• LayoutLM / Donut / TrOCR-based pipelines for more advanced custom setups

Recommendation by use case

• Fastest production-ready: Google Document AI, Azure Document Intelligence, or Amazon Textract
• Best enterprise workflow and customization: ABBYY
• Lowest cost / self-hosted: PaddleOCR plus your own table extraction

Important note

If you specifically need line items, make sure the tool supports:

• table detection
• key-value extraction
• confidence scores
• multi-page invoices
• post-processing for totals validation

If you tell me your environment—cloud preference, volume, budget, and whether invoices are mostly standard or messy—I can recommend the best one more specifically.