anyconvert

Python document conversion engine

anyconvert converts PDF documents into DOCX, PPTX, ODT, ODP, TXT, SVG, and HTML5 using only the Python standard library.

Highlights

• Dual Layout Engines:
  • Canvas Mode (--mode canvas): 1:1 pixel-accurate positioning. Renders exact coordinates, font families, embedded images, and vector artwork (native DrawingML custom geometries in DOCX and PPTX, SVG paths in ODT, and CSS/SVG canvas in HTML5). Ideal for flyers, forms, annotated PDFs, and presentation slides.
  • Flow Mode (--mode flow): Semantic reflowable document reconstruction. Automatically detects headings (H1–H6), bullet/numbered lists, tables, and reflowable paragraphs.
• Native Vector Graphics & DrawingML Engine:
  • High-precision PDF vector path parser decomposing cubic Bézier curves, subpaths, line caps, line joins, miter limits, and dash patterns.
  • Generates native OpenXML DrawingML custom geometries (<a:custGeom>) in DOCX and PPTX, native ODF <draw:path> in ODT/ODP, and standalone scalable vector graphics (.svg).
• Single-File HTML5:
  • Semantic reflowable output (<article>, <header>, <footer>, <table class="pure-table">, <ul class="flow-list">) and pixel-accurate canvas (position: absolute, inline <svg>, affine matrix transforms).
  • Truly portable single-file packaging with Base64 Data URI image embedding and zero-telemetry offline Content-Security-Policy (<meta http-equiv="Content-Security-Policy">).
• Comprehensive PDF Coverage:
  • Classic ASCII xref tables, xref streams (/Type /XRef), and object streams (/Type /ObjStm).
  • Standard PDF Security Revisions 2–6 (RC4 40–128 bit, AES-128, and AES-256 CBC).
  • Decompression filters (FlateDecode with PNG/TIFF predictors, LZWDecode, ASCII85Decode, ASCIIHexDecode, RunLengthDecode, CCITTFaxDecode, JBIG2Decode, JPXDecode).
  • Modern image support: In-memory PNG extraction (ISO/IEC 15948), JPEG, JPEG 2000, and native WebP (RIFF/VP8/VP8L/VP8X).
  • Full typography engine: Type 1, TrueType/OpenType SFNT tables, CFF/Type 2 CharStrings, CIDFonts, ToUnicode CMaps, Adobe Glyph List, and dynamic /Differences encodings.
• Standards-Compliant Packaging:
  • OPC (ISO/IEC 29500-2): Generates valid WordprocessingML (.docx) and PresentationML (.pptx) packages with complete [Content_Types].xml and .rels hierarchies.
  • ODF (ISO/IEC 26300): Generates compliant OpenDocument Text (.odt) and Presentation (.odp) archives with uncompressed mimetype headers at byte offset 0.
  • W3C SVG & HTML5: Standards-compliant vector graphics and self-contained HTML5 reports.

Supported Format Matrix

Known Issues

• Slide Orientation (ODP & PPTX):
  • If input PDF page aspect ratio is vertical/portrait, ODP and PPTX default slide dimensions currently scale to widescreen aspect ratios (resolving in Phase 12).

Installation

Install from PyPI:

pip install anyconvert

Or install in editable development mode:

git clone https://github.com/word-sys/anyconvert.git
cd anyconvert
pip install -e .

Command-Line Interface (CLI)

anyconvert provides a standalone command-line tool:

anyconvert [OPTIONS] INPUT_PDF

Options

CLI Examples

Single-File Responsive Web Document (HTML5 Flow):

anyconvert article.pdf -f html -m flow -o article.html

1:1 Pixel-Accurate Offline Web Report (HTML5 Canvas):

anyconvert flyer.pdf -f html -m canvas -o flyer.html

Standalone Scalable Vector Graphics Page (SVG):

anyconvert blueprint.pdf -f svg -o blueprint.svg

1:1 Pixel-Accurate Visual Export (DOCX):

anyconvert flyer.pdf -f docx -m canvas -o flyer.docx

1:1 Pixel-Accurate Visual Export (ODT):

anyconvert document.pdf -f odt -m canvas -o document.odt

Semantic Reflowable Conversion for Editing (DOCX):

anyconvert article.pdf -f docx -m flow -o article.docx

Convert PDF Slides to PowerPoint with Vectors (PPTX):

anyconvert presentation.pdf -f pptx -m canvas -o presentation.pptx

Convert Encrypted / Password-Protected PDF:

anyconvert secret.pdf -f docx -p "MyPassword123"

Python API

High-Level File Conversion (convert)

from anyconvert import convert, ConversionMode

# Accurate visual export to Word or LibreOffice Writer
convert(
    input_path="input.pdf",
    output_format="docx",
    output_path="output.docx",
    mode=ConversionMode.CANVAS,  # or "canvas"
)

# Semantic reflowable export to OpenDocument Text
convert(
    input_path="input.pdf",
    output_format="odt",
    output_path="output.odt",
    mode=ConversionMode.FLOW,  # or "flow"
)

In-Memory Byte Conversion (convert_bytes)

Ideal for web APIs, AWS Lambda, Cloud Run, and microservices:

from anyconvert import convert_bytes

pdf_bytes = open("document.pdf", "rb").read()

# Convert in-memory without touching disk
docx_bytes = convert_bytes(
    pdf_bytes=pdf_bytes,
    output_format="docx",
    mode="canvas",
)

Document Intermediate Representation (pdf_to_document_ir)

Inspect and manipulate document AST nodes before emitting:

from anyconvert import pdf_to_document_ir

doc_ir = pdf_to_document_ir("document.pdf", mode="canvas")

print(f"Total Pages: {len(doc_ir.pages)}")
print(f"Metadata: {doc_ir.metadata}")

for page in doc_ir.pages:
    print(f"Page {page.page_number} ({page.width}x{page.height} pt): {len(page.blocks)} blocks")

Architecture

[ Raw PDF Bytes / Stream ]
            │
            ▼
[ PDF Lexer & Parser ] ───► [ XRef Resolver & ObjStm Unpacker ]
                                    │
   ┌────────────────────────────────┼────────────────────────────────┐
   ▼                                ▼                                ▼
[ Decryption Engine ]    [ Decompression Filters ]        [ Typography Engine ]
(RC4 / AES-128 / AES-256) (Flate/LZW/ASCII85/PackBits)  (SFNT / CFF / Type 1 / CMap)
   │                                │                                │
   └────────────────────────────────┼────────────────────────────────┘
                                    │
                                    ▼
                    [ Content Stream Interpreter ]
                      (Graphics & Text Evaluator)
                                    │
                                    ▼
                 [ Spatial Indexing & Recursive XY-Cut ]
                                    │
                                    ▼
                      [ Semantic Feature Detectors ]
                     (Headings / Lists / Tables / Flow)
                                    │
                                    ▼
                         [ DocumentIR Builder ]
                                    │
       ┌─────────────────────┬──────┴──────────────┬─────────────────────┐
       ▼                     ▼                     ▼                     ▼
[ DOCX Emitter ]      [ PPTX Emitter ]      [ ODT / ODP / TXT ]   [ HTML5 / SVG ]
  (DrawingML)           (DrawingML)           (ISO/IEC 26300)      (Flow & Canvas)

Development & Verification

Running Tests

python3 -m pytest tests/

Static Type Checking

The codebase enforces 100% strict type safety (mypy --strict):

python3 -m mypy --strict src/ tests/

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for details on our architectural principles, development setup, code of conduct, and pull request guidelines.

License

GPL-3.0-or-later. See LICENSE for details.