Features
Text extraction from scanned documents and images
Handwriting recognition (cursive, mixed-content)
Complex table reconstruction with HTML markup (colspan/rowspan)
Form field extraction and layout analysis
Low-quality scan enhancement and OCR
Embedded image extraction from documents
Markdown output for easy downstream consumption
Structured JSON output via Document AI Playground
Batch processing for high-volume workloads
Cost-effective scaling ($2 per 1,000 pages)
50% discount for asynchronous batch API
Integrates with Mistral AI Studio for no-code parsing
Supports PDF and common image formats
Preserves document structure (headings, lists, tables)
Self-hosting option for data privacy
Pre-trained AI models for invoices, receipts, contracts
No-code custom field training
Real-time data validation rules
Batch processing with auto-file-upload
Multi-language document support
Handwriting extraction (limited)
Export to CSV, JSON, Excel
Role-based user permissions
Document classification by type
