Any-to-Markdown

What does Any-to-Markdown do?

Any-to-Markdown converts images and scanned PDFs into structured Markdown using AI-powered OCR. Unlike basic OCR tools, it understands complex document layouts and recognizing mathematical formulas (LaTeX), tables, figures, and text in their correct reading order.

It supports 80+ languages and handles academic papers, technical documents, invoices, and any document where layout matters.

Key features

• Formula recognition — Detects mathematical formulas and converts them to LaTeX
• Table extraction — Recognizes table structures and outputs them as Markdown tables
• Layout analysis — Understands document structure (titles, paragraphs, figures, captions)
• Multi-language OCR — Supports 80+ languages including Chinese, Japanese, Korean, Arabic, and more
• PDF support — Process multi-page PDF documents with page selection
• Multiple input methods — Upload files directly or provide URLs

Use cases

• Academic paper digitization — Convert scanned research papers with equations into editable Markdown/LaTeX
• Technical document processing — Parse engineering specs, datasheets, and manuals preserving tables and formulas
• Invoice and receipt parsing — Extract structured data from scanned financial documents
• Book digitization — Convert scanned book pages into searchable, editable text
• Data pipeline integration — Use the Apify API to automate document parsing in your workflows
• RAG / LLM preparation — Convert documents to Markdown for use as context in AI applications

How to use

1. Upload files or provide URLs to images (PNG, JPEG, BMP, WebP) or PDF documents
2. Select a recognition mode based on your content type
3. Choose your languages and configure options
4. Run the Actor and get structured Markdown output

Recognition modes

Mode	Best for	Description
Page (default)	General documents	Full layout analysis — detects titles, text, formulas, tables, and figures
Text + Formula	Math-heavy content	Mixed text and formula recognition without layout analysis
Formula Only	Equations	Pure mathematical formula to LaTeX conversion
Text Only	Simple documents	Standard OCR without formula detection
PDF	Multi-page docs	Processes PDF files with optional page selection

Supported file formats

Format	Extensions	Notes
PDF	.pdf	Multi-page documents, scanned or digital. Use pageNumbers to select specific pages.
PNG	.png	Recommended for screenshots and scanned documents
JPEG	.jpg, .jpeg	Photos, scans, camera captures
BMP	.bmp	Uncompressed bitmap images
TIFF	.tiff, .tif	High-quality scans, multi-layer images
WebP	.webp	Modern web image format
GIF	.gif	Static GIF images (first frame only)

Input

The Actor accepts uploaded files or URLs pointing to images or PDFs. All other parameters are optional with sensible defaults.

The most important input fields are:

• Upload Files — Drag and drop image or PDF files directly
• URLs — List of URLs pointing to documents to parse
• Recognition Mode — Select the type of content analysis (default: page)
• Languages — Comma-separated language codes, e.g. en,ch_sim (default: en,ch_sim)

See the Input Schema tab for the full list of configuration options including table recognition, formula detection, image resize settings, and output format.

Output

The Actor outputs results to the Dataset (JSON) and/or Key-Value Store (Markdown files), depending on the outputFormat setting.

Dataset output example

{
  "source": "https://example.com/document.png",
  "success": true,
  "recognitionMode": "page",
  "markdown": "## Introduction\n\nThe equation $E = mc^2$ describes...\n\n| Column A | Column B |\n|----------|----------|\n| Value 1  | Value 2  |",
  "format": "png",
  "sizeBytes": 245000,
  "processingTimeMs": 35000,
  "elements": [
    { "type": "TITLE", "text_preview": "Introduction" },
    { "type": "TEXT", "text_preview": "The equation..." },
    { "type": "FORMULA", "text_preview": "E = mc^2" },
    { "type": "TABLE", "text_preview": "| Column A |..." }
  ],
  "metadata": {
    "parsed_at": "2026-02-19T03:08:04.944383",
    "processing_time_ms": 35000,
    "file_size_bytes": 245000,
    "format": "PNG"
  }
}

Key-Value Store output

When using keyValueStore or both output format, each successfully parsed document is saved as a .md file in the Key-Value Store, ready to download or use in downstream workflows.

Tips to optimize costs

• Use Text Only mode for simple documents without formulas — it's faster
• Set Disable Formula Recognition to true if you don't need LaTeX output
• Use Page Numbers to process only specific pages of large PDFs
• Lower the Resized Shape value (e.g., 512) for faster processing at the cost of some accuracy

Supported languages

The Actor supports 80+ languages via its built-in text recognition engine. When using only English (en) or Simplified Chinese (ch_sim), a dedicated fast OCR engine is used automatically. For all other languages, the full multilingual engine is activated.

Provide multiple languages as comma-separated codes, e.g. en,ch_sim,ja

Common language codes

Code	Language	Code	Language
en	English	ch_sim	Chinese (Simplified)
ch_tra	Chinese (Traditional)	ja	Japanese
ko	Korean	vi	Vietnamese
fr	French	de	German
es	Spanish	pt	Portuguese
it	Italian	nl	Dutch
ru	Russian	uk	Ukrainian
ar	Arabic	fa	Persian (Farsi)
hi	Hindi	ta	Tamil
th	Thai	id	Indonesian
ms	Malay	tl	Filipino
tr	Turkish	pl	Polish
cs	Czech	ro	Romanian
hu	Hungarian	el	Greek
sv	Swedish	da	Danish
no	Norwegian	fi	Finnish
bn	Bengali	ur	Urdu
ne	Nepali	mn	Mongolian
my	Myanmar	ka	Georgian