PVPDF Vision
Se connecterCommencer gratuitement
PDF to JSON

Extract JSON from a PDF

Pull structured data out of any PDF as clean JSON — define your fields and let AI fill them.

Extract JSON from a PDF online

Instead of dumping the whole document as text, PDF Vision asks which fields matter — invoiceNumber, invoiceDate, totalAmount, partyA — and what type each one is: string, number, boolean, date, array or object. The AI then reads the PDF and returns exactly that shape as JSON, ready to paste into a request body, a database row or an automation step. Presets for invoices, contacts and contracts give you a starting schema. Extraction runs on our servers against a paid AI service, so this tool requires a Gold plan — a free account is refused on this endpoint.

How to extract JSON from a PDF

  1. Define your schema

    Add one row per field, name it in camelCase and set its type, or load the invoice, contact or contract preset.

  2. Upload the PDF

    Select the document to read; extraction runs on our servers under your Gold plan and takes longer on dense or multi-page files.

  3. Copy or download JSON

    Inspect the result in the tree view, then copy it to your clipboard or save it as a .json file.

Why use it

You control the schema

Field names and types are yours, so the response drops straight into an existing TypeScript interface or database column without writing a mapping layer in between.

Typed, not stringly typed

Declaring a value as number, boolean or date means totals come back as numbers and dates in a comparable form, instead of strings your code has to parse and validate.

Every run is saved

Past extractions stay in a history panel: reopen one to inspect its JSON in the tree view or download the file again, without re-uploading the PDF.

Frequently asked questions

How do I choose which fields to extract?
You list them yourself, one row at a time: a field name plus a type from string, number, boolean, date, array or object. Names must be valid identifiers, so write invoiceTotal rather than "invoice total".
Is there a starting point for common documents?
Yes, three presets fill the schema for you: invoice (number, date, vendor, total, currency), contact (full name, email, phone, company) and contract (title, both parties, start and end dates). Edit any row afterwards.
Does it work on scanned PDFs?
Not reliably. Extraction reads the text the PDF already contains, so a pure scan returns thin or empty fields. Run OCR PDF on it first to add a text layer, then come back and extract.
What if a field is not in the document?
AI extraction can only read what appears on the page, so a field the document never mentions cannot be filled. Treat the output as data to validate, especially figures that drive payments or accounting entries.
Do I need an account?
Yes, and a paid one: PDF to JSON requires a Gold plan. A free account is rejected on this tool, so signing up alone will not give you access. Extraction also runs server-side, which means the document is uploaded for processing rather than parsed inside your browser.
What can I do with the JSON afterwards?
Copy it from the tree view or download the file, then POST it to an API, insert it as a database row, or hand it to a spreadsheet import, accounting system or automation workflow.

Related tools