python-pptx

python-pptx is a Python library for creating, reading, and updating PowerPoint (.pptx) files without needing PowerPoint installed or licensed. It is commonly used to generate presentations from dynamic content such as database queries, analytics output, or JSON payloads, often in response to an HTTP request that returns a generated PPTX file.

The library models the OOXML slide format as a tree of Python objects — presentations, slides, shapes, text frames, tables, and charts — so consumers can build or edit decks with plain Python objects instead of manipulating XML directly. It can also analyze existing decks in a corpus, for example to extract text for search indexing or pull embedded images.

What You Get

• A Presentation object model covering slides, slide layouts, and slide masters
• Shape APIs for text boxes, images, tables, and native PowerPoint charts
• Read/write access to existing .pptx files for templating and bulk editing
• Fine-grained control over fonts, colors, alignment, and placeholders
• Support for embedding images (BMP, GIF, JPEG, PNG, TIFF, WMF) and chart data

Common Use Cases

• Generating PowerPoint reports from database queries or analytics pipelines
• Serving a dynamically-built .pptx file as an HTTP response
• Templating recurring decks (e.g. weekly status decks) by editing a base .pptx
• Extracting text and images from a corpus of presentations for search indexing or analysis
• Automating tedious manual slide production tasks

Under The Hood

Architecture — python-pptx models a .pptx file as an OPC (Open Packaging Conventions) package: pptx/opc/package.py handles the zip-based part/relationship graph, while a content_type_to_part_class_map in pptx/__init__.py dispatches each XML part (slide, slide layout, slide master, chart, image, core properties) to a dedicated Part subclass under pptx/parts/. User-facing objects (Presentation, Slide, Shape, Table, Chart) in pptx/presentation.py, pptx/slide.py, pptx/shapes/, pptx/table.py, and pptx/chart/ wrap these parts’ underlying lxml-parsed OOXML elements (pptx/oxml/), giving a Pythonic object layer over the raw XML tree while lazily writing changes back on save. Tech Stack — pure Python (94% of the codebase) with a thin lxml dependency for XML parsing, Pillow for image introspection, and XlsxWriter for embedded chart workbook data; packaging uses setuptools with pyproject.toml, and the project ships inline type stubs (py.typed, a typings/ directory) checked with pyright. Code Quality — the repo pairs a large unittest/pytest-based tests/ tree (83 test files) with a parallel features/ directory of Gherkin behavioral specs, giving both unit-level and end-to-end acceptance coverage; module boundaries are organized by OOXML concern (opc, oxml, parts, shapes, dml, enum) which keeps low-level XML manipulation isolated from the public object API. API Design — the public surface is a single Presentation(path) entry point returning an object graph navigated with attribute access (prs.slides, slide.shapes, shape.text_frame) rather than raw XML manipulation, which keeps common operations (add a slide, insert a table, set text) low-boilerplate, though deeper formatting control still requires understanding the underlying OOXML concepts the object model is built on.