AWS Textract goes beyond simple OCR to extract text, tables, forms, and structured data from scanned documents. With pricing that varies significantly based on which features you use, understanding the cost difference between DetectText and AnalyzeDocument is critical to controlling your document processing spend.
TL;DR: DetectText (basic OCR) costs $1.50 per 1,000 pages. AnalyzeDocument with Forms jumps to $50.00 per 1,000 pages. Tables cost $15.00 per 1,000 pages. Free tier includes 1,000 pages per month for 3 months. Use DetectText for simple OCR and reserve AnalyzeDocument for documents that actually need structured extraction.
DetectText Pricing (Basic OCR)
| Monthly Volume | Price per 1,000 Pages |
|---|---|
| First 1 million pages | $1.50 |
| Over 1 million pages | $0.60 |
DetectDocumentText extracts raw text and word-level confidence scores from document images. It returns lines and words with bounding box coordinates but does not identify form fields, table structures, or key-value pairs.
DetectText is the most cost-effective Textract feature and is suitable for simple document digitization, search indexing, and text archival workflows.
AnalyzeDocument Pricing
| Feature | Price per 1,000 Pages (First 1M) | Price per 1,000 Pages (Over 1M) |
|---|---|---|
| Forms (key-value pairs) | $50.00 | $40.00 |
| Tables | $15.00 | $10.00 |
| Queries | $15.00 | $10.00 |
| Signatures | $15.00 | $10.00 |
| Layout | $6.50 | $4.00 |
AnalyzeDocument adds structural understanding to text extraction. You can enable multiple features in a single API call. Each enabled feature is billed separately per page.
Feature Combination Costs
| Features Enabled | Cost per 1,000 Pages |
|---|---|
| Forms only | $50.00 |
| Tables only | $15.00 |
| Forms + Tables | $65.00 |
| Forms + Tables + Queries | $80.00 |
| All features | $101.50 |
A single page processed with Forms and Tables enabled costs $0.065 per page. At scale, this difference from the $0.0015 per page DetectText rate is over 40x more expensive.
Specialized API Pricing
AnalyzeExpense
| Volume Tier | Price per Page |
|---|---|
| First 1 million pages | $0.008 |
| Over 1 million pages | $0.004 |
AnalyzeExpense is purpose-built for receipts and invoices. It identifies vendor names, line items, prices, and totals. At $8.00 per 1,000 pages, it is significantly cheaper than using AnalyzeDocument with Forms ($50/1,000) for expense processing.
AnalyzeLending
| Component | Price per Page |
|---|---|
| Lending document classification and extraction | $0.007 |
AnalyzeLending is designed for mortgage and lending document processing. It automatically classifies document types (pay stubs, bank statements, tax forms) and extracts relevant fields. At $7.00 per 1,000 pages, it provides specialized extraction at a fraction of the generic AnalyzeDocument cost.
Free Tier
| Feature | Free Allowance | Duration |
|---|---|---|
| DetectText | 1,000 pages/month | 3 months |
| AnalyzeDocument | 1,000 pages/month | 3 months |
| AnalyzeExpense | 100 pages/month | 3 months |
| AnalyzeLending | 2,000 pages/month | 3 months |
The Textract free tier provides 3 months of free usage starting from your first API call. The free allowance applies independently to each API, meaning you can process 1,000 pages with DetectText and 1,000 pages with AnalyzeDocument in the same month at no charge.
Real-World Cost Examples
| Use Case | API Used | Monthly Volume | Monthly Cost |
|---|---|---|---|
| Document archival (OCR only) | DetectText | 50,000 pages | $75 |
| Invoice processing | AnalyzeExpense | 10,000 invoices | $80 |
| Contract extraction (forms) | AnalyzeDocument (Forms) | 5,000 pages | $250 |
| Tax form processing | AnalyzeDocument (Forms + Tables) | 20,000 pages | $1,300 |
| Mortgage processing | AnalyzeLending | 100,000 pages | $700 |
| Enterprise document pipeline | Mixed APIs | 500,000 pages | $4,500 |
Textract vs Alternatives
| Solution | Cost per 1,000 Pages (Simple OCR) | Structured Extraction |
|---|---|---|
| Textract DetectText | $1.50 | No |
| Textract AnalyzeDocument | $15-$50+ | Yes |
| Google Document AI | $1.50 (basic), $30 (forms) | Yes |
| Azure AI Document Intelligence | $1.50 (read), $50 (prebuilt) | Yes |
| Open-source (Tesseract) | Compute costs only | Limited |
Cost Optimization Tips
1. Use DetectText for Simple OCR
If you only need raw text extraction without form fields or table structures, use DetectText at $1.50 per 1,000 pages instead of AnalyzeDocument. This single decision can reduce costs by 90% or more for many document processing workflows.
2. Use AnalyzeExpense for Invoices and Receipts
AnalyzeExpense costs $8.00 per 1,000 pages compared to $50.00+ per 1,000 pages for AnalyzeDocument with Forms. For receipt and invoice processing, the specialized API is both cheaper and more accurate.
3. Enable Only Required Features
When calling AnalyzeDocument, only enable the features you need. Processing with Forms alone costs $50/1,000 pages. Adding Tables unnecessarily increases it to $65/1,000 pages. Evaluate each feature independently.
4. Pre-Process Documents Before Extraction
Clean up scanned documents (deskew, remove noise, enhance contrast) before sending to Textract. Higher quality inputs produce better results and reduce the need to reprocess pages. Store documents in Amazon S3 for direct integration with Textract async APIs.
5. Use Asynchronous APIs for Multi-Page Documents
The async APIs (StartDocumentTextDetection, StartDocumentAnalysis) process multi-page PDFs efficiently and publish results to SNS when complete. This avoids timeout issues and lets you process documents through Lambda-triggered pipelines at lower operational cost.
Related Guides
- AWS Bedrock Pricing Guide
- AWS Comprehend Pricing Guide
- AWS Lambda Pricing Guide
- AWS S3 Pricing Guide
FAQ
What counts as one page in Textract pricing?
One page is a single image (JPEG, PNG) or one page of a multi-page PDF document. If you submit a 10-page PDF to DetectText, you are billed for 10 pages ($0.015 total at the first-tier rate).
Can I process handwritten text with Textract?
Yes. Textract supports both printed and handwritten text extraction through DetectText and AnalyzeDocument. Handwritten text is included at the same price per page with no additional surcharge.
How does Textract compare to Amazon Rekognition text detection?
Rekognition DetectText is designed for text in photos (signs, labels, packaging) at $0.001 per image. Textract is designed for document pages and provides word-level confidence, table extraction, and form parsing. Use Rekognition for scene text and Textract for document processing.
Lower Your Textract Costs with Wring
Wring helps you access AWS credits and volume discounts to lower your Textract document processing costs. Through group buying power, Wring negotiates better rates so you pay less per page processed.
