Skip to main content

Amazon Textract

Amazon Textract FAQs

General

Open all

    Amazon Textract is a document analysis service that includes Amazon Textract Core and Amazon Textract Insights, the generative AI feature.

    Amazon Textract Core detects and extracts printed text, handwriting, structured data (such as fields of interest and their values), and tables from images and scans of documents. Its machine learning models have been trained on millions of documents so that virtually any document type you upload is automatically recognized and processed for text extraction. When information is extracted, the service returns a confidence score for each element it identifies so that you can make informed decisions about how to use the results. For instance, if you are extracting information from tax documents, you can set custom rules to flag any extracted information with a confidence score lower than 95%. All extracted data is also returned with location information, including a bounding box and a polygon that encompass each piece of data identified, so that you can quickly identify where a word or number appears on a document. You can access Textract Core through the AWS Management Console, the AWS SDKs, or the AWS Command Line Interface (AWS CLI).

    Amazon Textract Insights is a capability of Amazon Textract that builds on Textract Core's text and layout extraction. It uses generative AI to split multi-document files, classify each document, and extract the fields you define into structured data in a single asynchronous request, with a confidence score and location for every value. It can also return LLM-ready text, generate document summaries, and detect and redact personally identifiable information (PII). Textract Insights is available in the US East (N. Virginia) and US West (Oregon) Regions through the AWS SDKs and the AWS CLI.

    The most common use cases for Amazon Textract Core include:

    • Importing documents and forms into business applications
    • Creating smart search indexes
    • Building automated document processing workflows
    • Maintaining compliance in document archives
    • Extracting text for natural language processing (NLP)
    • Extracting text for document classification

    For Amazon Textract Insights, common use cases include:

    • Splitting, classifying, and extracting data from mixed document packets, such as loan files, claims files, and onboarding packages
    • Extracting the fields you define from documents such as invoices, tax forms, bank statements, pay stubs, and identity documents into structured records for downstream systems
    • Preparing documents for generative AI and retrieval-augmented generation (RAG) applications
    • Detecting and redacting personally identifiable information (PII) before extracted content reaches downstream systems

    For Amazon Textract Core: Amazon Textract detects printed text, including right-to-left and mixed-script documents, in more than 40 languages, including more than 30 European languages as well as Arabic, Chinese (Simplified and Traditional), Hebrew, Hindi, Japanese, and Korean. It also recognizes English handwriting. Every word and line comes with a bounding box and a confidence score, so each result can be traced back to the source page and relied on in automated workflows.

    Beyond text, Amazon Textract understands how a document is organized. Layout analysis identifies titles, headers, section headings, paragraphs, lists, figures, page numbers, and footers, and returns the content in its natural reading order, even on complex multi-column pages. Tables are extracted with their rows, columns, and merged cells intact in every supported language, including skewed or rotated tables. Signatures are detected wherever they appear on the page. For English, Spanish, Italian, Portuguese, French, and German documents, Amazon Textract also extracts forms as key-value pairs, along with checkboxes and other selection elements.

    Amazon Textract also extracts explicitly labeled data, implied data, and line items from almost any invoice or receipt in English, without templates or configuration. It can also extract specific or implied data, such as names and addresses, from English-language identity documents, such as U.S. passports and driver's licenses, without templates or configuration. Using Queries in English, you can extract specific data from a document regardless of how it's structured or laid out.

    For Amazon Textract Insights: Amazon Textract Insights extracts a document's text, layout elements, tables, and figures in reading order, and renders them as plain text, Markdown, or HTML, with tables also available as CSV. It also extracts the specific fields you define in a blueprint, such as single values, tables of repeating rows, or groups of related fields, and can generate a caption for each chart, graph, or image. Textract Insights supports English, Spanish, Italian, Portuguese, French, and German.

    Amazon Textract Core supports PDF, PNG, JPEG, and TIFF files. For synchronous APIs, you can submit documents either as an Amazon S3 object or as a byte array. For asynchronous APIs, you can submit Amazon S3 objects. If your document is already in one of these formats, don't convert or downsample it before uploading it to Amazon Textract.

    Amazon Textract Insights accepts PDF, PNG, JPEG, and TIFF files. You submit documents as Amazon S3 objects, and each request processes one input file as an asynchronous job. An input file can be up to 500 MB and 3,000 pages. Your input and output S3 buckets must be in the same Region as the request. Some features have lower page limits: custom extraction supports up to 20 pages per document unless you enable document splitting, and figure captioning supports up to 1,500 pages.

    To get started with Amazon Textract Core, choose the "Get Started with Amazon Textract" button on the Amazon Textract page. You must have an Amazon Web Services account; if you do not already have one, you will be prompted to create one during the process. After you sign in to your AWS account, try Amazon Textract Core with your own images or PDF documents in the Amazon Textract Management Console. You can also download the Amazon Textract SDKs to start creating your own applications. For more information, see the step-by-step Getting Started Guide.

    To get started with Amazon Textract Insights, use the AWS SDKs or the AWS CLI in the US East (N. Virginia) or US West (Oregon) Region. Create a project that defines what to extract, for example using a public W-2 blueprint, and then start a job on a document in your Amazon S3 bucket. When the job finishes, the results are written to your output bucket. For a step-by-step walkthrough, see the step-by-step Getting Started Guide.

    For Amazon Textract Core:

    • Detect Document Text. Performs OCR to detect printed text in a document.
    • Analyze Document. Detects printed text, handwriting, key-value pairs, tables, and other entities, along with their relationships and confidence scores, so extracted text keeps its intended structure. Use it to capture structured data from documents such as tax forms, financial reports, medical records, and loan applications.
      • Queries let you specify the data you need without worrying about how it's structured or laid out across different formats and versions of a document.
      • Custom Queries let you customize Queries to improve extraction accuracy on business-specific documents.
    • Analyze Expense. Extracts data from invoices and receipts, even when it isn't explicitly labeled. For example, it finds a vendor name that appears only in a logo, and line items, quantities, and prices that have no column headers. It returns normalized key names and column headers, so downstream applications can easily compare output from many documents.
    • Analyze ID. Understands the context of identity documents, such as U.S. passports and driver's licenses, without templates or configuration. Businesses that provide ID verification, and those in finance, healthcare, and insurance, can use it to automate account creation, appointment scheduling, employment applications, and more from a picture or scan of an ID.
    • Analyze Lending. Processes mortgage loan packages by classifying each page and routing it to the appropriate analysis.

    For details, see the Amazon Textract API reference.

    For Amazon Textract Insights:

    Textract Insights can replace multiple task-specific calls with a single asynchronous API:

    • StartDocumentInsights. Processes a document according to a reusable project, or with blueprints you pass in the request. The project defines what to extract and which capabilities to apply, such as document splitting, custom extraction with blueprints, and PII redaction.
    • GetDocumentInsights. Returns the job status and, when the job succeeds, the location of the results in your Amazon S3 bucket.
    • Project and blueprint operations. Create and manage projects and blueprints with operations such as CreateProject and CreateBlueprint.

    Analyze Document API has the following features – Forms, Tables, Queries, Custom Queries, Signatures, and Layout. You can use these features independently or use any combination of these features together. Use Forms to extract data such as key-value pairs (e.g., “First Name” and associated value “Jane Smith”). Use Tables to extract tabular or table data organized in columns and rows. Use Queries to specify the information you need from a document in the form of natural language questions (e.g., “What is the customer name?”) and receive the answer (e.g., “Jane Doe”) as part of the response. Use Custom Queries to customize the Queries features on business-specific documents. You can use Signatures to detect signatures on documents and use Layout to identify Layout elements in a document.

    In general, customers should try to ask a natural language question utilizing words from the document to construct a query. For detailed guidance on best practices for crafting Queries, see Best Practices for Queries.

    Queries are processed on a per page basis and information can be extracted using Queries via both synchronous and asynchronous operations. For synchronous operations, a maximum of 15 Queries per page is supported. For asynchronous operations, a maximum of 30 Queries per page is supported.

    For Amazon Textract Core: Amazon Textract uses machine learning to read virtually any type of document to extract printed text, handwriting, and structured information. To get the best results:

    • Make sure your document uses a supported language:
      • Printed text: Albanian, Arabic, Basque, Belarusian, Bosnian, Bulgarian, Chinese (Simplified and Traditional), Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Hebrew, Hindi, Hungarian, Icelandic, Italian, Japanese, Korean, Latvian, Lithuanian, Luxembourgish, Norwegian, Polish, Portuguese, Romanian, Russian, Serbian Cyrillic, Serbian Latin, Slovak, Slovenian, Spanish, Swedish, Turkish, Turkmen, and Ukrainian
      • Forms (key-value pairs and selection elements): English, French, German, Italian, Portuguese, and Spanish
      • Handwriting, invoices and receipts, identity documents, and Queries: English only
    • Provide as high quality an image as you can, ideally at least 150 DPI.
    • If your document is already in one of the file formats that Amazon Textract supports (PDF, PNG, JPEG, and TIFF), don't convert or downsample it before uploading it to Amazon Textract.
    • For table extraction, results are best when tables are visually separated from surrounding elements on the page, for example not overlaid on an image or complex pattern.

    You can get started with analyzing your own documents with Amazon Textract with just a few clicks in the Amazon Textract Management Console. If you have trouble achieving high accuracy with receipts, identification, or industrial diagrams, contact us at amazon-textract@amazon.com for assistance.

    For Amazon Textract Insights: To get the best results:

    • Make sure your document uses a supported language: English, Spanish, Italian, Portuguese, French, or German.
    • When your document type is fixed, use one blueprint per project. This skips the blueprint-matching step. If a project needs several blueprints, give each a descriptive name and description, and don't attach two blueprints for the same document type.
    • Write specific blueprint field descriptions. Include box or line numbers, expected formats, and common synonyms. For example, "Social security tax withheld (box 4)" works better than "tax withheld."
    • Use per-field confidence scores to send low-confidence values to human review.
    • For PII redaction, choose the specific entity types you need rather than all types. This reduces over-masking.
    • Enable only the capabilities and output formats you need. Each one adds processing time.

    A confidence score indicates how likely it is that a given prediction is correct. You can use confidence scores to set thresholds for your workflow; for example, automatically accepting high-confidence values and sending lower-confidence values to human review.

    For Amazon Textract Core: Amazon Textract returns a confidence score from 0 to 100 for each extracted element, including printed text, handwriting, and structured data, along with the element's location on the page as a bounding box and a polygon. This lets you review the score for each extracted entity and make informed decisions about how to use the results.

    For Amazon Textract Insights: Amazon Textract Insights returns confidence scores from 0 to 1. Each field you extract with a blueprint has its own score and location, so you can route individual fields to review rather than whole documents. Insights can also return a recognition score for each line and word of extracted text, when line- or word-level granularity is enabled, and a match score showing how well each document fits the blueprint applied to it.

    Yes.

    For Amazon Textract Core: All Amazon Textract Core operations are logged as CloudTrail events, including DetectDocumentText, AnalyzeDocument, AnalyzeExpense, AnalyzeID, and the asynchronous Start and Get operations.

    For Amazon Textract Insights: All Textract Insights operations are logged as CloudTrail events, including project and blueprint operations such as CreateProject and CreateBlueprint, and the StartDocumentInsights and GetDocumentInsights operations.

    For more information, see Logging Amazon Textract API Calls with AWS CloudTrail.

    Amazon Textract Core is available in the US East (N. Virginia), US East (Ohio), US West (Oregon), US West (N. California), AWS GovCloud (US-West), AWS GovCloud (US-East), Canada (Central), Europe (Ireland), Europe (London), Europe (Frankfurt), Europe (Paris), Europe (Spain), Europe (Zurich), Asia Pacific (Singapore), Asia Pacific (Sydney), Asia Pacific (Seoul), Asia Pacific (Mumbai), and Asia Pacific (Tokyo) Regions.

    Amazon Textract Insights is available in the US East (N. Virginia) and US West (Oregon) Regions.

    You can view and manage your Amazon Textract service quotas (formerly referred to as service limits) in the AWS Service Quotas console. To request an increase:

    1. Sign in to the AWS Console, navigate to the AWS Service Quotas console and select "Textract" under AWS services.
    2. Select the desired quota and click "Request Quota Increase" on the subsequent page.
    3. Enter the desired quota value and click "Request."

    You can estimate the quotas your use case needs with the Textract Service Quota calculator.

    For Amazon Textract Insights: Amazon Textract Insights has its own quotas, such as the number of concurrent jobs for each job type, and projects and blueprints per account. You can request increases to adjustable Insights quotas the same way outlined above. Some limits are fixed and can't be raised, such as the maximum blueprint schema size and the single fallback blueprint per project. For details, see Limits and quotas.

    We recommend the following approach to mitigate throttling:

    1. Implement retry logic. Follow the Error handling guidelines to configure retries for throttling errors.
    2. Configure exponential backoff and jitter. Adding exponential backoff and jitter to your retries improves throughput. See Error retries and exponential backoff in AWS.
    3. Smooth your traffic flow. Spiky traffic reduces throughput. To get maximum throughput for your allotted transactions per second (TPS), use a serverless queueing architecture or another mechanism to smooth traffic so it's more consistent.
    4. Start with samples that apply best practices. Try the IDP CDK Samples using CDK Constructs.
    5. Request a quota increase if needed. Use the Textract Service Quota calculator to estimate the quota requirements for your use case and submit a quota increase request from the AWS Service Quotas console.

    For Amazon Textract Insights: If you exceed the request rate quota, Amazon Textract Insights returns a ProvisionedThroughputExceededException. Retry with exponential backoff. If you reach the quota for concurrent jobs, it returns a LimitExceededException. Retry after running jobs finish, or request a quota increase if you need more throughput. When you retry StartDocumentInsights, pass a ClientRequestToken so the retry returns the original job instead of starting a duplicate. Because jobs are asynchronous, design your client to poll GetDocumentInsights with a backoff rather than blocking.

    Amazon Textract Core APIs return building blocks, such as words, lines, tables, and key-value pairs, or a fixed set of fields for supported document types. Amazon Textract Insights returns ready-to-use output in a single asynchronous request: Markdown, HTML, and plain text representations of each document, with tables also available as CSV, and exactly the fields you define in a blueprint, for any document type. It can also split files that contain several documents, generate summaries and figure captions, and detect and redact PII. Use Amazon Textract Core when you need a synchronous response, cell-level table detail, or raw text only. Forms as key-value pairs, Queries, and Custom Queries adapters are available in Amazon Textract Core only, so if you need those alongside Amazon Textract Insights capabilities, you can use both.

    No. Your existing Amazon Textract Core API calls, including DetectDocumentText, AnalyzeDocument, AnalyzeExpense, and AnalyzeID, keep the same endpoints, request formats, response schemas, and pricing, and any Custom Queries adapters you've trained keep working. Amazon Textract Insights adds new APIs, such as StartDocumentInsights and GetDocumentInsights, when you need capabilities such as blueprints, document splitting, LLM-ready output, or PII redaction. Amazon Textract Insights is available in the US East (N. Virginia) and US West (Oregon) Regions. In all other Regions, you continue to use Amazon Textract Core as you do today.

Billing

Open all

    An image (PNG or JPEG) counts as a single page. For PDF and TIFF files, each page in the document counts as a page processed.

    Refer to the Amazon Textract pricing page to learn more about pricing.

    Amazon Textract charges you based on the number of pages and images processed. For more information, visit the pricing page.

    Yes. As part of the AWS Free Tier, you can get started with Amazon Textract Core for free. The Free Tier lasts for three months, and new AWS customers can analyze up to:

    Detect Document Text API: 1,000 pages per month
    Analyze Document API:

    • 1,000 Pages per month when using Signatures only
    • 100 Pages per month when using Forms, Tables, and Layout features
    • 100 pages per month each for Queries, Forms + Queries, Tables + Queries, Forms + Tables + Queries
    • There is no free-tier for Custom Queries

    Analyze Expense API: 100 pages per month
    Analyze ID API: 100 pages per month
    Analyze Lending API: 2,000 pages per month

Data Privacy

Open all

    Amazon Textract Core and Amazon Textract Insights may store and use document and image inputs processed by the service solely to provide and maintain the service and to improve and develop the quality of Amazon Textract and other Amazon machine-learning/artificial-intelligence technologies. Use of your content is necessary for continuous improvement of your Amazon Textract customer experience, including the development and training of related technologies. We do not use any personally identifiable information that may be contained in your content to target products, services or marketing to you or your end users. Your trust, privacy, and the security of your content are our highest priority and we implement appropriate and sophisticated technical and physical controls, including encryption at rest and in transit, designed to prevent unauthorized access to, or disclosure of, your content and ensure that our use complies with our commitments to you. Please see Data Privacy FAQ for more information. You may opt out of having your document and image inputs used to improve or develop the quality of Amazon Textract and other Amazon machine-learning/artificial-intelligence technologies using an AWS Organizations opt-out policy. For information about how to opt out, see Managing AI services opt-out policy.

    For Amazon Textract Core: Any content processed by Amazon Textract Core is encrypted and stored at rest in the AWS Region where you are using Amazon Textract. Unless you opt out as provided below, some portion of content processed by Amazon Textract may be stored in another AWS Region solely in connection with the continuous improvement and development of your Amazon Textract customer experience and other Amazon machine-learning/artificial-intelligence technologies. You can request deletion of document and image inputs associated with your account by contacting AWS Support. Your trust, privacy, and the security of your content are our highest priority and we implement appropriate and sophisticated technical and physical controls, including encryption at rest and in transit, designed to prevent unauthorized access to, or disclosure of, your content and ensure that our use complies with our commitments to you. Please see Data Privacy FAQ for more information. Your content will not be stored in another AWS Region if you opt out of having your content used to improve and develop the quality of Amazon Textract and other Amazon machine-learning/artificial-intelligence technologies. For information about how to opt out, see Managing AI services opt-out policy.

    For Amazon Textract Insights: Amazon Textract Insights uses cross-Region inference. Your data remains in the geography where the request originates, but during inference, your content may be processed in another AWS Region within the same geography, where it may be stored for abuse detection for up to 30 days. For example, requests made in US East (N. Virginia) or US West (Oregon) may be processed in US East (N. Virginia), US East (Ohio), US West (N. California), or US West (Oregon). All data is encrypted in transit across Amazon's network. Your Amazon S3 input and output locations remain in the Region you call, and a customer managed AWS KMS key remains scoped to that Region.

    Yes. You can request deletion of document and image inputs associated with your account by contacting AWS Support. Deleting image and document inputs may degrade your Amazon Textract experience.

    Yes. You always retain ownership of your content, and we will only use your content with your consent.

    Any content used for generating adapters is processed internally within Amazon Textract for the duration of the training. The content is encrypted at rest and in transit. The content is stored and processed in the AWS Region where you are training the adapter, and is deleted once training completes. Please see Data Protection in Amazon Textract for more information.

    Yes, AWS has expanded its HIPAA compliance program to include Amazon Textract as a HIPAA eligible service. If you have an executed Business Associate Agreement (BAA) with AWS, you can use Amazon Textract to extract text including protected health information (PHI) from images.

    Learn more about HIPAA Compliance

    Amazon Textract is HIPAA-eligible and compliant with PCI, ISO, and SOC. For more information, please visit AWS Artifact in the AWS Management Console, or visit AWS Services in Scope by Compliance Program. Textract also supports Amazon Virtual Private Cloud (Amazon VPC) endpoints via AWS PrivateLink, enabling customers to securely initiate API calls to Amazon Textract from within their VPC and avoid using the public internet.