PptxConverter

Convert PowerPoint PPTX presentations to Markdown with images, tables, and charts

Overview

The PptxConverter class converts Microsoft PowerPoint .pptx files to Markdown. It extracts slide content including text, images (with AI captioning support), tables, charts, and speaker notes.

Dependencies

pip

pip install markitdown[pptx]

Requires: python-pptx

Accepted Formats

MIME Types · list

• application/vnd.openxmlformats-officedocument.presentationml*

Extensions · list

• .pptx

Class Definition

class PptxConverter(DocumentConverter):
    """Converts PPTX files to Markdown.
    
    Supports headings, tables and images with alt text.
    """

Constructor

def __init__(self):
    super().__init__()
    self._html_converter = HtmlConverter()

Methods

accepts()

def accepts(
    file_stream: BinaryIO,
    stream_info: StreamInfo,
    **kwargs: Any,
) -> bool

Returns True if the file has a .pptx extension or PowerPoint MIME type.

convert()

def convert(
    file_stream: BinaryIO,
    stream_info: StreamInfo,
    **kwargs: Any,
) -> DocumentConverterResult

Converts a PPTX presentation to Markdown.

Parameters:

file_stream · BinaryIO · required

Binary stream of the PPTX file

stream_info · StreamInfo · required

Metadata about the file

llm_client · OpenAI client

OpenAI-compatible client for AI image captioning

llm_model · str

Model to use for image captioning (e.g., “gpt-4o”, “gpt-4-vision-preview”)

llm_prompt · str

Custom prompt for image captioning. Overrides default prompt.

keep_data_uris · bool · default: "False"

If True, embeds images as base64 data URIs. If False, uses placeholder filenames.

Returns: DocumentConverterResult with converted Markdown

Raises: MissingDependencyException if python-pptx is not installed

Features

Slide Elements

• Titles - Converted to Markdown H1 headings
• Text - Regular text frames preserved
• Images - Extracted with AI-generated descriptions or alt text
• Tables - Converted to Markdown tables
• Charts - Extracted as Markdown tables with data
• Notes - Speaker notes included under “Notes” section
• Grouped Shapes - Recursively processed

Image Handling

Images can have descriptions from multiple sources:

1. AI Caption (if llm_client provided) - Generated description using vision model
2. Embedded Alt Text - Description from PowerPoint
3. Shape Name - Fallback to shape name

Descriptions are combined and sanitized for Markdown.

Chart Extraction

Charts are converted to Markdown tables:

### Chart: Sales by Quarter

| Category | Q1 | Q2 | Q3 | Q4 |
|----------|----|----|----|----|  
| Product A | 100 | 150 | 200 | 180 |
| Product B | 80 | 90 | 120 | 140 |

Unsupported chart types show [unsupported chart].

Example Usage

Basic Conversion

from markitdown.converters import PptxConverter
from markitdown._stream_info import StreamInfo

converter = PptxConverter()

with open("presentation.pptx", "rb") as f:
    stream_info = StreamInfo(extension=".pptx")
    result = converter.convert(f, stream_info)
    print(result.markdown)

With AI Image Captioning

from openai import OpenAI

client = OpenAI(api_key="your-api-key")

with open("presentation.pptx", "rb") as f:
    stream_info = StreamInfo(extension=".pptx")
    result = converter.convert(
        f, 
        stream_info,
        llm_client=client,
        llm_model="gpt-4o",
        llm_prompt="Describe this slide image in detail."
    )
    print(result.markdown)

With Base64 Image Embedding

with open("presentation.pptx", "rb") as f:
    stream_info = StreamInfo(extension=".pptx")
    result = converter.convert(
        f,
        stream_info,
        keep_data_uris=True  # Embed images as data URIs
    )
    print(result.markdown)

Output Example

<!-- Slide number: 1 -->
# Welcome to Our Product

Introduction to the new features

![A screenshot showing the product dashboard with analytics](Picture1.jpg)

### Notes:
Remember to emphasize the key benefits

<!-- Slide number: 2 -->
# Sales Data

### Chart: Quarterly Revenue

| Category | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| Revenue | 100000 | 120000 | 150000 | 180000 |

Shape Processing Order

Shapes are processed in visual order (top-to-bottom, left-to-right) based on their position on the slide:

sorted_shapes = sorted(
    slide.shapes,
    key=lambda x: (x.top, x.left)
)

Implementation Details

Source Location

~/workspace/source/packages/markitdown/src/markitdown/converters/_pptx_converter.py:34

Helper Methods

• _is_picture() - Detects picture shapes
• _is_table() - Detects table shapes
• _convert_table_to_markdown() - Converts tables via HTML intermediary
• _convert_chart_to_markdown() - Extracts chart data as tables

Slide Structure

Each slide includes:

1. Slide number comment
2. Title (if present)
3. Content (shapes in visual order)
4. Notes section (if present)

Limitations

• Animations and transitions not preserved
• SmartArt converted to text only
• Some complex chart types show as [unsupported chart]
• Video/audio embedded content not extracted
• Layout and visual styling information lost