From high-speed OCR for templated documents to AI-powered Vision API for complex scanned content, Nutrient provides the right extraction technology for your specific needs. Process digitally born PDFs, scanned images, forms, and unstructured documents with industry-leading accuracy.
Process thousands of documents with classic OCR — 120+ languages, minimal overhead, optimized for invoice batches and search indexing.
Pull out structured data, including tables with cell boundaries, handwritten text, mathematical equations, and key-value pairs using Vision API’s local ICR or cloud VLM-enhanced processing.
Create WCAG-compliant alt text and image captions with Claude, OpenAI, or local AI models for complete privacy control.
Process documents entirely on-premises for HIPAA compliance and air-gapped environments, or connect to cloud VLMs when accuracy takes priority.
AI DOCUMENT PROCESSING
Automatically recognize document types and extract structured data using AI Document Processing’s hybrid approach — combining proven key-value pair technology with large language models for unmatched accuracy across 100+ file types.
Automatically identify invoices, resumes, passports, purchase orders, and more using AI-powered classification.
Start immediately with 10 built-in templates for common documents, or build custom templates for specialized workflows.
Combine large language models with proven key-value pair extraction for best-in-class accuracy on unstructured documents.
Seamlessly integrate into .NET 8.0+ applications with multithreaded performance for high-volume processing.
OCR
Extract text from scanned documents at speed with classic OCR — available across all platforms with 120+ languages. Optimized for invoice batches, search indexing, and high-volume processing where speed takes priority over semantic analysis.
Process documents with OCR on Web, .NET, Python, Java, iOS, and Android — consistent text extraction across your entire stack.
Recognize text in 120+ languages — the most comprehensive language coverage for global document processing.
Minimal computational overhead for high-volume scenarios like invoice batches, receipt scanning, and search indexing.
Automatic deskew, noise reduction, and orientation detection — plus zonal OCR for targeted field extraction.
VISION API
Analyze document structure with Vision API’s intelligent content recognition. Choose local ICR for privacy-sensitive processing, or connect to Claude/OpenAI for VLM-enhanced accuracy — available for Python and Java.
Analyze document layout with semantic element detection — identify headings, paragraphs, tables, lists, equations, and hierarchical relationships.
Run entirely offline with local AI models for HIPAA, SOC 2, and air-gapped environments — no external API calls required.
Optionally connect to Claude or OpenAI for VLM-enhanced ICR with improved table cell boundaries and classification confidence.
Generate WCAG-compliant alt text and image captions using Claude, OpenAI, or local AI models for complete privacy control.
COMPARE
Compare OCR, Vision API, and AI Document Processing to find the best fit for your document type, workflow, and accuracy requirements.
| OCR | Vision API | AI Document Processing | |
|---|---|---|---|
| Best for document type | Digitally born PDFs with fixed templates — invoices,
receipts, forms where layout is consistent | Scanned documents or non-digitally born content — handwritten
forms, research papers, variable layouts | Documents requiring classification and extraction — invoices,
resumes, passports with varying formats |
| Platform availability | All platforms: Web, .NET, Python, Java, iOS, Android,
Document Engine | Python and Java SDKs only | .NET 8.0+ only |
| Processing mode | Text extraction only — converts images to searchable text
with minimal overhead | Intelligent content recognition (ICR) — local AI models or
optional cloud VLM enhancement | Hybrid LLM + key-value pair — on-premises classification and
extraction |
| Key capabilities | 120+ languages, zonal OCR, preprocessing (deskew, noise
reduction), throughput-optimized | Table extraction with cell boundaries, handwriting
recognition, equation detection (LaTeX), document structure
analysis | 10 preconfigured templates, automated document
classification, structured data extraction, custom template
builder |
| Template requirements | Best with fixed templates — use Document Engine + Viewer if
you control PDF creation | No templates required — handles variable layouts and
unstructured content | Works with or without templates — 10 built-in templates adapt
to format variations |
| When to use | You control PDF creation or the document follows a fixed
template — high-volume batches where speed is priority | Content isn’t digitally born or no fixed template exists —
need semantic understanding of document structure | Need classification and extraction in one step — processing
mixed document types at scale |
| Language support | 120+ languages (Document Engine) — most comprehensive
coverage in industry | Multilingual support via AI models — handles mixed-language
documents | Multilingual via LLM — optimized for common business
documents |
| Deployment options | Fully on-premises — no external API calls required | Local ICR for privacy or cloud VLM for accuracy — flexibility
to balance compliance vs. performance | Fully on-premises with .NET runtime — meets HIPAA, SOC 2,
air-gapped requirements |
| Ideal use cases | Invoice batches, receipt scanning, search indexing,
high-volume text extraction | Forms with handwriting, research papers with equations,
complex tables, scanned documents | Automated invoice processing, resume parsing, passport
extraction, mixed document workflows |
PROVEN AT SCALE
Replaced paper and email with Nutrient Workflow to automate multilevel approvals across six Latin American offices, processing 236 asset requests.
Renders multipage PDFs and signature tags with Nutrient, keeping 200 million users in 188 countries moving at the speed of eSignature.
Empowers 34,000 pilots to view, annotate, and sign 90‑page flight releases on iPad using Nutrient iOS SDK, saving minutes — and money — on every flight.
Nutrient’s Document Web Services (DWS) platform offers cloud-native APIs that support every stage of the document lifecycle — from rendering a single PDF in the browser, to high-volume processing and automation.
Handle advanced, headless document workflows from the cloud — generate, convert, extract, add watermarks, and more with a processing API built for scale.
Deliver rich, interactive document experiences directly in the browser — render PDFs, annotate, fill forms, and collect signatures with a cloud-hosted viewer API.
Choose based on your document type and workflow needs. Use OCR for high-throughput text extraction from scanned documents with minimal overhead — ideal for invoice batches and search indexing. Use Vision API when you need to understand document structure, extract tables with cell boundaries, recognize handwriting, or process mathematical equations. Use AI Document Processing when you need automated document classification combined with data extraction using preconfigured templates — perfect for invoices, resumes, and passports.
OCR is available across all platforms: Web, .NET, Python, Java, iOS, Android, and Document Engine — making it the most versatile option for cross-platform text extraction. Vision API is currently available for Python and Java SDKs, offering intelligent content recognition with local or cloud processing. AI Document Processing is exclusively available for .NET 8.0+ applications, providing LLM-powered classification and extraction.
Nutrient’s Document Engine OCR supports 120+ languages — the most comprehensive language coverage in the industry. This includes complex scripts like Arabic, Chinese, Japanese, Korean, Thai, Hebrew, and Cyrillic alphabets. Language packs can be selectively loaded to optimize memory usage and processing speed based on your application’s specific requirements.
Yes. Vision API’s intelligent content recognition (ICR) can extract tables with complete structural information, including cell boundaries, row and column relationships, and hierarchical data. It also recognizes handwritten text, mathematical equations (outputting LaTeX format), and document layout elements like headings, paragraphs, and lists. You can use local ICR for privacy-sensitive processing or connect to Claude/OpenAI for VLM-enhanced accuracy.
Local ICR runs entirely on-premises using local AI models — perfect for HIPAA compliance, SOC 2 requirements, and air-gapped environments with no external API calls. Cloud VLM-enhanced ICR optionally connects to Claude or OpenAI for improved table cell boundary detection and higher classification confidence on complex documents. Both modes are available in Python and Java, giving you complete flexibility to balance privacy requirements against accuracy needs.
AI Document Processing includes 10 preconfigured templates for common document types (invoices, resumes, passports, purchase orders, etc.) that work immediately without configuration. You can also build custom templates for specialized workflows using the template builder. The hybrid LLM + key-value pair approach adapts to document variations within each category. A single invoice template handles multiple layouts — no need to create separate templates for every format.
Yes. OCR and Vision API with local ICR both support completely offline, on-premises processing with no external API calls — meeting HIPAA, SOC 2, and air-gapped environment requirements. Vision API can use local AI models (LM Studio, Ollama, vLLM) for image description generation while maintaining full data privacy. AI Document Processing also runs entirely on-premises within your .NET application infrastructure.
All three technologies support 100+ file types, including PDFs (digitally born and scanned), image formats (TIFF, JPG, PNG, BMP, GIF, HEIF), Office documents (Word, Excel, PowerPoint), email formats (EML, MSG), web formats (HTML, MHTML), and text formats (RTF, TXT). Vision API and OCR are optimized for image-based documents, while AI Document Processing excels at processing any document type that requires classification and structured extraction.
Vision API provides the most advanced handwriting recognition using intelligent content recognition with optional VLM enhancement; it’s capable of handling cursive writing, mixed printed/handwritten text, and mathematical equations. OCR includes basic handwriting recognition suitable for forms and structured documents with clearly written text. For production handwriting workflows, Vision API’s ICR mode is recommended for the highest accuracy.
Yes. Vision API includes image description generation using Claude, OpenAI, or local AI models — creating WCAG-compliant alt text and detailed image captions for accessibility compliance. You can choose cloud-based models (Claude, OpenAI) for highest quality descriptions, or use local AI models (LM Studio, Ollama) to keep sensitive document images completely private while still generating accessible descriptions.
Start a free trial with watermark-limited access to OCR, Vision API, and AI Document Processing across all supported platforms. The trial includes full documentation, code samples, and developer support to evaluate features before purchasing. For extended testing or enterprise demos, contact Sales for a temporary production license key.
Explore platform-specific guides for Web OCR, Python Vision API, or AI Document Processing. Each guide includes complete code samples, API references, and step-by-step tutorials. You can also try the interactive Web Playground to test OCR capabilities in your browser without any installation.
Learn how to integrate real-time PDF editing into your app with Nutrient SDKs. Enhance document collaboration with instant annotations and synchronization, and seamless multi-user editing.
Explore strategies to save and sync annotations using external files for better efficiency.
We talk about how the layers feature of Instant can be used for construction site visits.
FOR DEVELOPERS
Nutrient provides three specialized technologies for document data extraction — OCR for high-speed text extraction, Vision API for intelligent document understanding, and AI Document Processing for automated classification and extraction. Each technology is optimized for specific use cases and document types, giving you the flexibility to choose the right tool for your workflow.
Choose the right technology based on your specific requirements:
Each technology supports different platforms and deployment models:
Balance privacy requirements against accuracy needs with flexible processing modes:
Each technology provides specialized capabilities for different extraction needs:
Nutrient’s data extraction platform provides unique advantages over competitors with fragmented offerings. Instead of choosing between basic OCR libraries or expensive cloud-only services, you get three specialized technologies that work together — offering cross-platform consistency, flexible deployment models, and the industry’s most comprehensive language support (120+ languages). Complete documentation, code samples, and developer support are included with every license, and all technologies support on-premises deployment for compliance with HIPAA, SOC 2, and air-gapped requirements.
