New: page-level confidence routing is live// document parsing api

Any PDF in. Structured Markdown/JSON out. Pay per page.

Scanned files, phone photos, academic papers, business reports — DocUnwrap parses them all into clean Markdown or JSON at ~$0.002 a page. Roughly 90% cheaper than AWS Textract for structured extraction.

100 pages free every month — no credit card.

quarterly-report.pdf
Page 1 of 3
title
table
$1,245.00
MarkdownJSONBlocks (12)
Q1 Financial Summary
MetricQ4Q1
Revenue$1,120$1,245
Margin31%34%
Pages1824
~$0.002
per page · base OCR
90%
cheaper than Textract structured extraction
100
free credits every month
// what it parses

One API for every document shape.

From fax-grade scans to dense financial tables — each type below is parsed end-to-end in the live demo.

Scanned documents

Fax-grade scans and photocopies, with a confidence score on every page.

Phone photos

Skewed, glare-struck photos of receipts, forms and whiteboards.

Academic papers

Two-column layouts, equations as LaTeX, references kept intact.

Financial reports

Dense numeric tables survive as clean Markdown tables, ready for analysis.

Invoices & receipts

Line items and totals come back as structured JSON with bounding boxes.

Multilingual & mixed-script

English, Chinese, Japanese, Korean and major European languages — mixed in one document.

// how it works

From PDF to pipeline-ready data in three steps.

Routing, retries and confidence handling happen between the lines — your side is one HTTP call.

01

Submit the PDF

POST the file or a URL with your API key. Born-digital PDFs take a zero-cost fast path and return in seconds.

POST /v1/parse
02

We parse & route

Each page runs on your chosen model tier; weak pages are automatically re-parsed by stronger degradation-tuned models.

multi-model routing
03

Consume the result

Poll the task or get a webhook callback — clean Markdown, structured JSON and a confidence score on every page.

webhook · poll
// docunwrap vs the incumbents

Structured extraction at a tenth of the incumbent price.

Everyone on this list parses PDFs. The difference is what comes back — and what each page costs you.

DocUnwrapAWS TextractMistral OCRAzure Document Intelligence
Base OCR price~$0.002/page$1.50 / 1k pages$2–4 / 1k pages$1.50 / 1k pages
Structured extraction price~$0.005/page$65 / 1k pages$2–4 / 1k pages$10–30 / 1k pages
Real-world degraded documentsStrongFairWeakFair
Table accuracyHighMediumMediumHigh
Model choiceTransparent & selectableBlack boxBlack boxBlack box
Async batch processingNativeLimitedLimitedLimited
// why teams switch

Built for real documents, priced for real volume.

Reads what others give up on

Phone photos, faxes, low-quality scans. Pages that come back weak are re-parsed by stronger degradation-tuned models before you ever see them.

90%+ cheaper structured extraction

Structured JSON at ~$0.005 a page — $5 per thousand pages, not $65. The savings compound with every document in your corpus.

Pricing you can compute

Per-page rates and overage published up front. No seat math, no sales call standing between you and the calculator.

The model is not a black box

Steer the accuracy/cost trade-off per request via an API parameter — and page-level confidence tells you which results to trust.

Async is the primary API

Batch queues, webhook callbacks, 100 concurrent tasks on day one. Bulk PDF processing is the core scenario, not an afterthought.

// what the api does

PDF to Markdown

Heading hierarchy, lists, tables and LaTeX formulas come through intact — output you can drop straight into RAG or an LLM context.

PDF to JSON

Structured blocks with bounding boxes, so every extracted value traces back to its exact region on the page.

Scanned & photo OCR

Phone photos, faxes, skewed low-quality scans. Weak pages are automatically re-routed to stronger degradation-tuned models.

Async batch processing

Queue large multi-page PDFs and get webhook callbacks. A 50-page batch typically completes in under a minute.

Digital-PDF fast path

Born-digital PDFs skip GPU inference entirely — results come back near-instantly and cost almost nothing to serve.

Table extraction

Complex tables — merged cells, multi-page spans — extracted as clean Markdown or HTML tables.

Page-level confidence

Per-page confidence scores flag what needs human review and drive automatic re-routing to stronger models.

Multilingual documents

English, Chinese, Japanese, Korean and the major European languages — mixed-script documents included.

// integrate

From upload to structured output in minutes.

One REST endpoint. No SDK required — an API key and a URL are all it takes.

  • Text-native PDFs return in seconds on a zero-cost fast path.
  • Async batch processing with webhook callbacks for large files.
  • Transparent per-page pricing — pay only for what you parse.
parse — docunwrap api
$ curl https://api.docunwrap.com/v1/parse \
  -H "Authorization: Bearer $DocUnwrap_API_KEY" \
  -F file=@quarterly-report.pdf
24 pages · queued · est. $0.12
markdown · tables · confidence 0.97
{ "format": "markdown", "pages": 24, "markdown": "# Quarterly Report…" }
$
// who builds on docunwrap

Built for teams whose pipeline starts with a document.

RAG · knowledge bases

AI application developers

Feed embeddings from clean Markdown instead of OCR soup — scanned papers and messy reports included — priced so indexing your whole corpus stays boring.

invoices · contracts · reports

SMB document automation

Photographed invoices and scanned contracts come back as structured JSON with bounding boxes — no human re-typing fields between the scanner and your ledger.

arXiv · LaTeX · tables

Academic & research tooling

Formulas as LaTeX, tables kept intact, page-level confidence for review queues — parse entire arXiv categories overnight with the batch API.

// questions, answered

Frequently asked questions

What is DocUnwrap?

DocUnwrap is a document parsing API. It converts any PDF — scanned documents, phone photos, academic papers, business reports — into structured Markdown or JSON through a single HTTP API, priced per page.

How much does it cost?

Basic OCR is about $0.002 per page and structured extraction about $0.005 per page. The Free plan includes 100 pages every month; paid plans start at $19/month for 4,750 credits, with overage from $0.003 per page.

Does it work on scanned or phone-photo PDFs?

Yes — that is the hard part we focus on. Low-quality scans, faxes and phone photos are parsed with page-level confidence scores, so you know exactly which pages need a second look.

What output formats do I get?

Markdown that preserves heading hierarchy, lists, tables and LaTeX formulas, or structured JSON with bounding boxes for every element — ready for RAG pipelines and data extraction.

Can I process large documents in bulk?

Async batch processing is a first-class citizen: submit a multi-hundred-page PDF, then poll the task or receive a webhook callback when parsing finishes. Native digital PDFs take a zero-cost fast path and return in seconds.

Is it cheaper than AWS Textract?

For structured extraction, roughly 90% cheaper: about $5 per 1,000 pages versus $65 per 1,000 pages on AWS Textract — with transparent per-page pricing and no hidden fees.

Parse your first 100 pages free.

Create an account, grab an API key, and make your first call in minutes. 100 pages a month free — no credit card.

$curl -F file=@doc.pdf -H "Authorization: Bearer $DocUnwrap_API_KEY" https://api.docunwrap.com/v1/parseStart free