U1-OCR
Recognize IDs, parse documents, extract data
Intelligent ID and document parsing, one-click key data extraction
U1-OCR: Recognize IDs, parse documents, extract data
U1-OCR is an intelligent document parsing and extraction model that moves beyond traditional OCR character recognition and upgrades from simply reading text to understanding documents and extracting key information. It handles ID recognition, layout restoration, and information extraction in one workflow, turning unstructured documents into clean, usable data across office files, IDs and receipts, and complex reports while greatly reducing manual entry and verification.
99+%
Content recognition accuracy
50+
Languages supported
100+
Document types covered
<1s
Single-page info extraction
On authoritative KIE benchmarks Nanonets-KIE and CC-OCR-KIE, U1-OCR reaches industry-leading SOTA scores of 93.4 and 94.86, outperforming mainstream multimodal OCR and general multimodal foundation models. It also achieves a SOTA-level 94.63 on the authoritative document parsing leaderboard OmniDocBench V1.5.
Key strengths
Key strengths
Free your hands from tedious work
With end-to-end intelligent processing, it automatically classifies files and filters key information so you can stop organizing folders and typing fields by hand and spend time on higher-value work.
Simple to start, easy for anyone
Individual users can upload files and use the full workflow instantly without specialized training, while enterprise users can flexibly customize features and connect systems quickly through standardized APIs.
Restore the original layout faithfully
Even with complex layouts and varied formats, it preserves the source document structure so outputs stay clean and organized without layout confusion, shifted content, or formatting loss.
All-format compatible, ready anytime
Whether the input is a casually captured image, a professional scan, or a mainstream office file, it uploads and parses smoothly so you can handle documents anywhere with less friction.
Flexible enough for diverse content
It goes beyond standard OCR limits to recognize handwriting, seals, handwritten annotations, and special symbols accurately, opening up more practical document scenarios.
Technical highlights
Technical highlightsDeep visual-semantic fusion
It does more than recognize text pixels by combining structural visual perception with textual semantic understanding to truly understand document layout logic and content meaning.
Adaptive restoration for irregular layouts
Specialized optimization handles skewed shots, creased pages, and non-standard layouts with automatic perspective correction and strong restoration beyond generic OCR approaches.
Full-stack intelligent understanding
A one-stop solution for document classification, layout restoration, content interpretation, and key extraction, handling everything from organizing files to capturing core information intelligently.
Normalized processing for heterogeneous inputs
It adapts to original photos, HD scans, complex layout documents, and blurry recaptures, producing unified structured output with strong material compatibility.
Use cases
Use casesFast ID data entry
Capture IDs, passports, bank cards, and similar documents to extract information in one click and avoid manual typing.
Smarter invoice reimbursement
Automatically recognize digital and paper invoices and extract amounts, dates, and headers for easier expense submission.
Convert handwritten notes to digital files
Turn class notes, meeting notes, and handwritten lists into searchable, editable text from a photo.
Extract multilingual materials with ease
Recognize and extract information from foreign-language materials, notes, and screenshots in one click for more efficient reading and organization.
Capabilities
CapabilitiesIntelligent document classification:
Powered by OCR 3.0 cognition, it automatically identifies document types and classifies them accurately across office and business documents, with JSON Schema support for custom categories.
General information extraction:
Using OCR 3.0 semantic capabilities, it extracts times, amounts, organizations, and other key fields automatically without predefined templates, reducing manual work in common business scenarios.
Custom Schema extraction:
Define target fields, formats, and rules with JSON Schema to capture specific business information precisely and improve processing efficiency and accuracy.
High-precision parsing for complex layouts:
It understands document hierarchy, mixed media, and sectional structure, optimizes irregular table parsing, restores table data completely, and parses professional report, ledger, and statement layouts accurately.
Recognition for unconventional complex content:
It adapts to non-standard documents and accurately recognizes handwriting, seals, annotations, code, and special symbols, reducing misses and errors common in traditional OCR.
Flexible pricing, tailored solutions, and private deployment
U1-OCR
Recognize IDs, parse documents, extract data
Intelligent ID and document parsing, one-click key data extraction
U1-OCR: Recognize IDs, parse documents, extract data
U1-OCR is an intelligent document parsing and extraction model that moves beyond traditional OCR character recognition and upgrades from simply reading text to understanding documents and extracting key information. It handles ID recognition, layout restoration, and information extraction in one workflow, turning unstructured documents into clean, usable data across office files, IDs and receipts, and complex reports while greatly reducing manual entry and verification.
Content recognition accuracy
Languages supported
Document types covered
Single-page info extraction
On authoritative KIE benchmarks Nanonets-KIE and CC-OCR-KIE, U1-OCR reaches industry-leading SOTA scores of 93.4 and 94.86, outperforming mainstream multimodal OCR and general multimodal foundation models. It also achieves a SOTA-level 94.63 on the authoritative document parsing leaderboard OmniDocBench V1.5.
Key strengths
Free your hands from tedious work
With end-to-end intelligent processing, it automatically classifies files and filters key information so you can stop organizing folders and typing fields by hand and spend time on higher-value work.
Simple to start, easy for anyone
Individual users can upload files and use the full workflow instantly without specialized training, while enterprise users can flexibly customize features and connect systems quickly through standardized APIs.
Restore the original layout faithfully
Even with complex layouts and varied formats, it preserves the source document structure so outputs stay clean and organized without layout confusion, shifted content, or formatting loss.
All-format compatible, ready anytime
Whether the input is a casually captured image, a professional scan, or a mainstream office file, it uploads and parses smoothly so you can handle documents anywhere with less friction.
Flexible enough for diverse content
It goes beyond standard OCR limits to recognize handwriting, seals, handwritten annotations, and special symbols accurately, opening up more practical document scenarios.
Technical highlights
Deep visual-semantic fusion
It does more than recognize text pixels by combining structural visual perception with textual semantic understanding to truly understand document layout logic and content meaning.
Adaptive restoration for irregular layouts
Specialized optimization handles skewed shots, creased pages, and non-standard layouts with automatic perspective correction and strong restoration beyond generic OCR approaches.
Full-stack intelligent understanding
A one-stop solution for document classification, layout restoration, content interpretation, and key extraction, handling everything from organizing files to capturing core information intelligently.
Normalized processing for heterogeneous inputs
It adapts to original photos, HD scans, complex layout documents, and blurry recaptures, producing unified structured output with strong material compatibility.
Use cases
Fast ID data entry
Capture IDs, passports, bank cards, and similar documents to extract information in one click and avoid manual typing.
Smarter invoice reimbursement
Automatically recognize digital and paper invoices and extract amounts, dates, and headers for easier expense submission.
Convert handwritten notes to digital files
Turn class notes, meeting notes, and handwritten lists into searchable, editable text from a photo.
Extract multilingual materials with ease
Recognize and extract information from foreign-language materials, notes, and screenshots in one click for more efficient reading and organization.
Capabilities
Intelligent document classification:
Powered by OCR 3.0 cognition, it automatically identifies document types and classifies them accurately across office and business documents, with JSON Schema support for custom categories.
General information extraction:
Using OCR 3.0 semantic capabilities, it extracts times, amounts, organizations, and other key fields automatically without predefined templates, reducing manual work in common business scenarios.
Custom Schema extraction:
Define target fields, formats, and rules with JSON Schema to capture specific business information precisely and improve processing efficiency and accuracy.
High-precision parsing for complex layouts:
It understands document hierarchy, mixed media, and sectional structure, optimizes irregular table parsing, restores table data completely, and parses professional report, ledger, and statement layouts accurately.
Recognition for unconventional complex content:
It adapts to non-standard documents and accurately recognizes handwriting, seals, annotations, code, and special symbols, reducing misses and errors common in traditional OCR.
