Amazon Textract

Textract turns images of documents into structured data. A developer sends a scanned file or PDF, and the service returns the text it finds along with its layout: it identifies tables and preserves their rows and columns, detects key-value pairs in forms such as a name field and its value, and recognizes handwriting. This is much more useful than plain text recognition, which would return an unordered block of characters.

Specialized features target common documents. Analyze ID reads passports and driver’s licenses, Analyze Expense extracts line items and totals from invoices and receipts, and lending-specific features process mortgage documents. The Queries feature lets you ask a question like what is the invoice total, and get the answer from the document. Results integrate with other AWS services, such as Comprehend, Bedrock, and Step Functions, for further analysis.

Textract is priced per page, with different rates for different features, and a free tier for the first months. Accuracy is good on clean documents but can drop with poor scans, unusual layouts, or handwriting. Sensitive documents require attention to security configuration. Alternatives include Google Document AI, Azure Document Intelligence, ABBYY, and open-source tools such as Tesseract, which is free but less capable on structure. Textract suits teams on AWS that process forms and invoices.