Docling v2.129.0
🇺🇸 English EN 🇩🇪 Deutsch DE 🇯🇵 日本語 JA 🇨🇳 简体中文 ZH 🇫🇷 Français FR 🇪🇸 Español ES
GitHub Get Started
Reference

Docling Supported Formats

What Docling can read and what it can export. Based on the official supported-formats list — always check the official page for the current set.

Format Category OCR Tables Notes Command
PDF
.pdf		 Documents optional yes Advanced PDF parsing with layout, reading order and table understanding.
  • Use --no-ocr for digital PDFs and --ocr-mode full_page for scans.
  • The most feature-complete input: layout, reading order and tables.
 docling convert file.pdf --to md
Word
.docx, .doc		 Documents no yes Word documents with headings, lists and tables.
  • Headings, lists and tables map closely to Markdown.
 docling convert file.docx --to md
OpenDocument Text
.odt		 Documents no yes OpenDocument text documents.
  • Ideal for documents created in LibreOffice Writer.
 docling convert file.odt --to md
Apple Pages
.pages		 Documents no yes Apple Pages documents (Pages 5+ and iWork '09).
  • Supports Pages 5+ and iWork '09 files.
 docling convert file.pages --to md
Markdown
.md, .qmd, .Rmd		 Documents no yes Plain Markdown and Markdown supersets.
  • Round-trips Markdown and its supersets (qmd, Rmd).
 docling convert file.md --to md
AsciiDoc
.adoc, .asciidoc		 Documents no yes AsciiDoc text documents.
  • Good for technical documentation written in AsciiDoc.
 docling convert file.adoc --to md
LaTeX
.tex, .latex		 Documents no yes LaTeX source documents.
  • Convert .tex source, not the compiled PDF.
 docling convert file.tex --to md
Plain text
.txt, .text		 Documents no no Plain-text files.
  • Plain text carries no structure to recover.
 docling convert file.txt --to md
PowerPoint
.pptx, .ppt		 Presentations no yes Slides with layout, text boxes and speaker notes.
  • Speaker notes are captured alongside the slide text.
 docling convert file.pptx --to md
OpenDocument Presentation
.odp		 Presentations no yes OpenDocument presentations.
  • Ideal for LibreOffice Impress decks.
 docling convert file.odp --to md
Excel
.xlsx, .xls		 Spreadsheets no yes Workbooks converted to tables per sheet.
  • Each sheet becomes its own structured table.
 docling convert file.xlsx --to md
OpenDocument Spreadsheet
.ods		 Spreadsheets no yes OpenDocument spreadsheets.
  • Ideal for LibreOffice Calc workbooks.
 docling convert file.ods --to md
CSV
.csv		 Spreadsheets no yes Comma-separated values.
  • Parsed as a single table.
 docling convert file.csv --to md
HTML
.html, .htm, .xhtml		 Web no yes Web pages and saved HTML.
  • Scripts and styles are ignored; visible content is kept.
 docling convert file.html --to md
EPUB
.epub		 Web no yes E-books in EPUB format.
  • Chapters follow the book's spine order.
 docling convert file.epub --to md
WebVTT
.vtt		 Web no no Web Video Text Tracks (subtitles/captions).
  • Timed captions become structured text.
 docling convert file.vtt --to md
Images
.png, .jpg, .jpeg, .tiff		 Images yes partial Scanned pages and image files; OCR required for text.
  • OCR is required for text; use --ocr-mode full_page.
 docling convert scan.png --to md
Email
.eml, .msg		 Email no yes Email message files.
  • Headers, body and attachments are separated.
 docling convert mail.eml --to md
XBRL
.xml (XBRL)		 XML no yes eXtensible Business Reporting Language financial reports.
  • Financial tables are preserved.
 docling convert report.xml --to md
USPTO XML
.xml (USPTO)		 XML no yes USPTO patent XML documents.
  • Patent XML with sections and claims.
 docling convert patent.xml --to md
JATS XML
.xml (JATS)		 XML no yes JATS journal article XML.
  • Journal article XML with structured sections.
 docling convert article.xml --to md
Audio
.wav, .mp3		 Audio / Video no no Speech transcribed with the ASR pipeline.
  • Needs the asr extra; uses Whisper.
 docling convert audio.mp3 --pipeline asr --to md
Video
.mp4, .avi, .mov, .mkv, .webm		 Audio / Video no no Video parsed into an ASR transcript with representative keyframes.
  • Produces an ASR transcript plus representative keyframes.
 docling convert video.mp4 --pipeline asr --to md
ToolConfig generator →Build CLI & Python snippets. ReferenceOCR engines →Compare engines and platform support. RecipesExamples →Task-oriented CLI and Python recipes. ReferenceCommand finder →Copy-pasteable commands per format.

Verified with Docling v2.129.0 · Last checked 2026-09-22 · Official source

1
Pipeline

How Docling converts a document

Docling reads 23 input formats across eight families and exports nine output formats. This guide explains how conversion works, how to read the table and how to choose the right output.

Every input flows through the same pipeline: the file is parsed into a single unified DoclingDocument, then exported to the format you choose.

  • Layout analysis detects columns, headers, footers and reading order.
  • TableFormer reconstructs tables into Markdown or HTML matrices.
  • OCR is applied only when a page has no text layer, or when you force it.
  • The result is one DoclingDocument you can export many ways.
2
Reference

How to read the table

The table above lists every supported input with its extensions, category and a ready-to-run command.

  • OCR: whether OCR applies to this input (optional, yes or no).
  • Tables: whether table structure is recovered (yes, partial or no).
  • Notes and tips: what to watch out for with this format.
  • Command: copy it and replace the file name with your own.
3
Output

Choosing the right output

The input format does not limit your export; pick the output that fits your next step.

ExportWhat you getBest for
Markdown (md)Readable text with headings and tablesDocs, notes and RAG text.
JSON (json)Lossless DoclingDocument with bounding boxesCustom pipelines and tooling.
HTML (html)Tables and layout as HTMLWeb previews and email.
DocTags (doctags)Compact token-style markupModel input.
DocLang (doclang)Docling's declarative document languageRound-tripping and advanced workflows.
WebVTT (vtt)Timed subtitles and captionsAudio and video transcripts.
YAML (yaml)YAML representation of the documentConfig-friendly structured output.
LaTeX (latex)LaTeX sourceScientific and academic publishing.
Chunks (chunks)Chunked JSON for vector storesRetrieval-augmented generation.
4
OCR & tables

OCR and tables by format

OCR only applies to image-based inputs. Tables are recovered for every structured format.

  • Images and scanned PDFs need OCR; digital PDFs usually do not.
  • Office, web and XML inputs already carry structure, so OCR is off by default.
  • Audio and video use the ASR pipeline instead of OCR.
  • Choose an OCR engine on the OCR engines page.
5
Limits

When a format is not supported

If a file is not in the table, Docling cannot parse it directly.

  • Convert it to PDF, DOCX or HTML first, then parse that.
  • Remove password protection from encrypted PDFs before converting.
  • Test with a fresh copy if a file may be corrupted.
  • See the troubleshooting page for conversion errors.
6
Performance

Performance and privacy

Conversion runs locally, and only the models you enable consume time.

  • Disable OCR for digital documents; it is the biggest speed factor.
  • Use --table-mode fast when approximate tables are acceptable.
  • Prefer a GPU for large batches of PDFs or images.
  • Nothing is uploaded; files never leave your machine.
7
FAQ

Frequently asked questions

Which formats can Docling read?
23 input formats across documents, spreadsheets, presentations, web, images, email, XML and audio/video. The table above lists them all.
What output formats can Docling write?
Markdown, JSON, HTML, DocTags, DocLang, WebVTT, YAML, LaTeX and chunked JSON.
Does Docling need OCR for every PDF?
No. OCR is optional and only needed for scans or PDFs without a text layer. Use --no-ocr for digital PDFs.
Can Docling read encrypted PDFs?
No. Remove password protection before converting.
Does Docling keep tables?
Yes. TableFormer reconstructs tables for structured inputs and PDFs; images and scans depend on OCR quality.
What if my format is not listed?
Convert it to PDF, DOCX or HTML first, then run Docling on that.