Flagship Product

TrueTrace

Extract. Ground. Verify.

TrueTrace is Roysa's document understanding engine for extracting structured data from scanned documents while preserving full traceability to the original source.

See It In Action

Your documents,
intelligently extracted

Watch as TrueTrace identifies, extracts, and grounds every field to its exact location on the page. Full traceability, zero black boxes.

Try TrueTrace
TrueTrace
Extracting...
warranty_deed.pdf Page 1 of 3
GRANTOR
GRANTEE
DATE
Extraction Results
GRANTOR
"John Smith"
98% p.1
GRANTEE
"ABC Holdings LLC"
96% p.1
EFFECTIVE_DATE
"2024-03-15"
99% p.1
PROPERTY_ADDRESS

How It Works

1

Upload

Upload PDFs or images of your scanned documents

2

Define Keys

Specify the fields you want to extract

3

Extract & Ground

AI extracts values and anchors them to source locations

4

Review

View annotations and confidence scores

5

Export

Export structured, auditable results

Core Capabilities

Multi-page Support

Process documents with any number of pages seamlessly

Key-Conditioned Extraction

Extract exactly what you specify, nothing more

Bounding Boxes

Visual annotations with precise page references

Confidence Scoring

Know exactly how certain each extraction is

Validation Hooks

Integrate custom validation rules and checks

Flexible Export

JSON, CSV, or integrate via API

Core Engine

Multimodal Document AI

Our models combine vision, layout understanding, and text reasoning to process documents the way humans do — considering visual structure, not just extracted text.

  • Vision + layout + text reasoning
  • Works beyond clean digital PDFs
  • Understands document structure
  • Handles complex layouts
👁️ Vision
📐 Layout
📝 Text
AI
Visual Intelligence

Grounded Extraction

Every extracted value is spatially aligned to its source location on the document. This isn't post-processing — it's how our models fundamentally operate.

  • Spatial alignment between values and document pixels
  • Robust to noise, skew, and low-quality scans
  • Precise bounding box coordinates
  • Page-level references
Document
Value 1
Value 2

Output Format

Every extraction includes complete context for verification

extraction_result.json
{
  "extractions": [
    {
      "key": "GRANTOR",
      "value": "John Smith",
      "bounding_box": [120, 340, 280, 365],
      "page": 1,
      "confidence": 0.98,
      "evidence": "...recorded by John Smith..."
    },
    {
      "key": "EFFECTIVE_DATE",
      "value": "2024-03-15",
      "bounding_box": [450, 520, 560, 545],
      "page": 1,
      "confidence": 0.95,
      "evidence": "...effective March 15, 2024..."
    }
  ]
}

Technical Specifications

Input Formats

  • PDF (scanned & digital)
  • JPEG, PNG, TIFF
  • Multi-page documents

Output Formats

Performance

  • Sub-second per page
  • Batch processing support
  • Horizontal scalability

Quality

  • Handles 72-600 DPI
  • Skew correction
  • Noise resilience

Deployment Options

Choose the deployment model that fits your security requirements

☁️

Cloud

Fully managed service hosted on secure cloud infrastructure. Quick setup with enterprise-grade security.

Available Now
🏢

Private Cloud

Dedicated infrastructure in your preferred cloud provider. Enhanced isolation and compliance.

Available
🔒

On-Premises

Full deployment within your own data center. Maximum control for air-gapped environments.

Coming Soon

Use Cases

Forms & Applications

Government forms, permit applications, registration documents

Invoices & Receipts

Financial documents requiring accurate data extraction

Engineering Drawings

Technical specifications and blueprint details

Inspection Reports

Compliance and quality assurance documentation

Legacy Archives

Historical scanned documents and records

Legal Documents

Contracts, agreements, and court filings

Experience TrueTrace

See how grounded extraction transforms your document workflow.