Output Formats

PDFPipe supports 10 output formats across three categories. Set the format field in your convert request to choose one, and optionally pages (for example "1-5" or "last") to limit which pages are included.

What the extractor does and does not do. The engine is pdf.js text extraction (via pdf-parse). It returns each page's text and the PDF's Info dictionary (Title, Author, Producer, CreationDate, and so on - whichever keys the file has). It does not perform OCR, so scanned PDFs return empty text. It does not detect tables, headings, lists, columns, fonts, or coordinates. If you need any of those, run the per-page text through your own parser or an LLM.

Inline delivery (returnMethod: "inline"): Text-oriented extraction formats (json, text, markdown, xml, csv) are usually the best fit for embedding in the JSON API response. Encoded and image formats (base64, binary, png, jpg, webp) are returned as Base64 strings when inline; the response includes contentEncoding: "base64" and a matching contentType (e.g. image/png). Very large outputs may fall back to a presigned URL with returnMethodFallback and returnMethodFallbackReason.

Format	Value	Category	Content-Type	Tiers
JSON	`json`	Extraction	`application/json`	All tiers
Text	`text`	Extraction	`text/plain`	All tiers
Markdown	`markdown`	Extraction	`text/markdown`	Starter+
XML	`xml`	Extraction	`application/xml`	Starter+
CSV	`csv`	Extraction	`text/csv`	Starter+
Base64	`base64`	Encoded	`text/plain`	Starter+
Binary	`binary`	Encoded	`application/pdf`	Starter+
PNG	`png`	Image	`image/png`	Starter+
JPG	`jpg`	Image	`image/jpeg`	Starter+
WebP	`webp`	Image	`image/webp`	Starter+

Extraction Formats

Extraction formats read the text layer of the PDF and return it per page, plus the PDF's metadata. They work on digital-born PDFs with a text layer; scanned or image-only PDFs return empty text because there is no OCR. None of them detect tables, headings, or layout.

JSON

format: "json"

Page-by-page text plus the PDF's own metadata (its Info dictionary) and the page count. Each page is { pageNumber, text }. There are no tables, coordinates, fonts, or bounding boxes.

Example JSON output

{
  "pages": [
    {
      "pageNumber": 1,
      "text": "Invoice #2026-0142\nDate: February 15, 2026\n\nItem  Qty  Price\nAPI Credits  1000  $49.00..."
    },
    {
      "pageNumber": 2,
      "text": "Terms and conditions..."
    }
  ],
  "metadata": {
    "Title": "Invoice #2026-0142",
    "Author": "Acme Corp",
    "Producer": "Acrobat Distiller 21.0",
    "CreationDate": "D:20260215100000Z"
  },
  "totalPages": 2,
  "extractedAt": "2026-08-23T12:00:01.840Z"
}

Common use cases

Data extraction pipelines that need page boundaries
LLM / RAG document ingestion
Reading the PDF's title, author, and creation date

Text

format: "text"

Plain text extraction. All pages concatenated, in reading order as stored in the PDF, with blank lines between pages. No structural metadata.

Example Text output

Invoice #2026-0142
Date: February 15, 2026

Item          Qty    Price
API Credits   1000   $49.00
Priority      1      $29.00

Total: $78.00


Terms and conditions
Payment is due within 30 days...

Common use cases

Full-text search indexing
Simple text processing
Content previews

Markdown

format: "markdown"

A top-level heading from the PDF's Title (or "Document" if it has none), then a "## Page N" heading followed by that page's plain text. Headings, lists, and tables inside the PDF are not detected; the page text is passed through as-is.

Example Markdown output

# Invoice #2026-0142

## Page 1

Invoice #2026-0142
Date: February 15, 2026

Item          Qty    Price
API Credits   1000   $49.00

## Page 2

Terms and conditions
Payment is due within 30 days...

Common use cases

LLM prompt context with page boundaries as headings
Chunking documents by page for RAG pipelines
Content migration

XML

format: "xml"

A <document> with a <metadata> block (one element per PDF Info key, plus <totalPages>) and a <pages> block with one <page number="N"><text>...</text></page> per page. Useful for systems that consume XML natively.

Example XML output

<?xml version="1.0" encoding="UTF-8"?>
<document>
  <metadata>
    <Title>Invoice #2026-0142</Title>
    <Author>Acme Corp</Author>
    <totalPages>2</totalPages>
  </metadata>
  <pages>
    <page number="1">
      <text>Invoice #2026-0142
Date: February 15, 2026...</text>
    </page>
    <page number="2">
      <text>Terms and conditions...</text>
    </page>
  </pages>
</document>

Common use cases

Enterprise system integrations
XSLT transformation pipelines
Legacy system compatibility

CSV

format: "csv"

Two columns, page_number and text_content, with one row per page. Text is quoted and escaped per RFC 4180. This is not table extraction: tables inside the PDF are not detected.

Example CSV output

page_number,text_content
1,"Invoice #2026-0142
Date: February 15, 2026

Item          Qty    Price
API Credits   1000   $49.00"
2,"Terms and conditions
Payment is due within 30 days..."

Common use cases

Loading page text into a spreadsheet or database, one row per page
Feeding per-page text into tools that only accept CSV
Quick page-count and text-length audits

Encoded Formats

Encoded formats return the raw PDF file content in a transport-friendly encoding. Useful when you need the original file rather than extracted text - for example to archive an attachment PDF that was captured by the headless browser.

Base64

format: "base64"

The raw PDF file content encoded as a Base64 string. Useful when you need the original PDF bytes embedded in a JSON payload or email.

Example Base64 output

JVBERi0xLjQKMSAwIG9iago8PAov
VHlwZSAvQ2F0YWxvZwovUGFnZXMg
MiAwIFIKPj4KZW5kb2JqCjIgMCAo...

Common use cases

Embedding PDFs in API responses
Email attachments
Systems that require Base64 input

Binary

format: "binary"

The raw PDF file bytes, exactly as fetched. The result URL serves the original PDF as a binary download; inline delivery base64-encodes it with contentEncoding: "base64".

Example Binary output

(Binary PDF data - download via the presigned result URL)

Common use cases

PDF archival and storage
Re-serving downloaded attachment PDFs
Proxying PDFs through your own system

Image Formats

Image formats render a single page - the first page of the PDF, or the first page in your pages selection - as a raster image. Useful for previews and thumbnails. To get a different page, set pages to that page number.

PNG

format: "png"

A rasterised image of the first page (or the first page in your pages selection) in PNG format. Lossless, best for documents with text. One image per request, not one per page.

Example PNG output

(PNG image data - download via the presigned result URL)

Common use cases

Document thumbnails and cover previews
Visual comparison and auditing
Feeding a page image to your own OCR or vision model

JPG

format: "jpg"

A rasterised image of the first page (or the first page in your pages selection) in JPEG format at quality 85. Smaller than PNG with lossy compression.

Example JPG output

(JPEG image data - download via the presigned result URL)

Common use cases

Web thumbnails where file size matters
Social media previews
Quick visual previews

WebP

format: "webp"

A rasterised image of the first page (or the first page in your pages selection) in WebP format at quality 85. Best balance of quality and file size for web display.

Example WebP output

(WebP image data - download via the presigned result URL)

Common use cases

Web applications optimised for performance
Mobile-friendly document previews
Progressive web apps