Convert PDF forms
into clean JSON
Upload a PDF, extract its form fields, review AI labels, and produce a validated Schema V3 structure ready for downstream automation.
Drag & drop your PDF
or click to browse
Upload target
Form filler
Open completed forms and download filled PDFs.
Platform Demo
How CM-PDF works
CM-PDF is a human-in-the-loop pipeline for turning messy PDF forms into reliable structured data. AI does the heavy extraction work, while reviewers keep control over labels, widget ownership, table structure, and the final database update.
Upload a PDF form
The platform stores the source PDF, detects duplicates, and prepares it for extraction.
Extract PDF widgets
AcroForm fields are detected with page, type, bounds, and nearby OCR text for reviewer context.
Link widgets to labels
A vision model proposes human-readable labels, then reviewers correct missed or ambiguous fields.
Build Schema V3
The form becomes clean JSON: sections, text, fields, groups, irregular tables, and repeat groups.
Review and update DB
Validation blocks duplicate, unknown, or missing widget ownership before saving the final output.
Live workflow preview
Stage 3 Form Builder
{
"schema_version": 3,
"title": "Patient Intake",
"pages": [{
"page": 1,
"items": [
{ "kind": "section",
"label": "Applicant" },
{ "kind": "field",
"label": "First name",
"input_type": "text",
"widgets": [
{ "widget_id": "page1-widget1" }
] },
{ "kind": "field",
"label": "Consent?",
"input_type": "radio",
"choices": [
{ "label": "Yes",
"widget_id": "page1-widget3" },
{ "label": "No",
"widget_id": "page1-widget4" }
] }
]
}]
}Precise Highlighting
Review every extracted PDF widget against the original page.
Smart Workspace
Move through extraction, linking, and schema review in one flow.
Universal Export
Save validated Schema V3 output for automation and training data.
Supports PDF files only