Microsoft PowerPoint (Open XML)

Robert Gaskins, Dennis Austin, and Thomas Rudkin built the original PowerPoint at Forethought, Inc. under the working name "Presenter." PowerPoint 1.0 shipped on April 20, 1987 for Macintosh. Microsoft acquired Forethought three months later for $14 million — its first significant acquisition. By the late 1990s, PowerPoint held approximately 95% worldwide market share in presentation software. The shift from binary PPT to XML-based PPTX arrived with Office 2007, driven by interoperability demands and the standardization push that produced ECMA-376.

Internal ZIP structure

Renaming any .pptx file to .zip and extracting it reveals a standardized directory tree. A python-pptx-generated test file with two slides contains 40 ZIP entries totaling 96987 bytes uncompressed (23636 bytes compressed — a 75.6% compression ratio):

presentation.zip/
├── [Content_Types].xml          ← MIME manifest for every part
├── _rels/.rels                   ← Root relationships (entry point)
├── docProps/
│   ├── core.xml                  ← Dublin Core metadata (author, dates)
│   ├── app.xml                   ← Application metadata (slide count)
│   └── thumbnail.jpeg            ← Preview image
└── ppt/
    ├── presentation.xml          ← Slide order, dimensions, master refs
    ├── slides/slide1.xml         ← Individual slide content
    ├── slideLayouts/             ← Layout templates (11 default)
    ├── slideMasters/             ← Master slide definitions
    ├── theme/theme1.xml          ← Color schemes, fonts, effects
    └── media/                    ← Embedded images, video, audio

The [Content_Types].xml file at the root declares the MIME content type of every part. It uses <Default> elements to map file extensions to types and <Override> elements for specific part paths. This is how a ZIP-based viewer distinguishes a PPTX from a DOCX or XLSX — all three share the same 50 4B 03 04 ("PK") magic bytes, but [Content_Types].xml declares application/vnd.openxmlformats-officedocument.presentationml.presentation.main+xml for the main presentation part.

The OPC relationship chain

_rels/.rels is the first file any OPC consumer reads. It contains <Relationship> elements mapping IDs to target parts. For PPTX, the root relationship type http://schemas.openxmlformats.org/officeDocument/2006/relationships/officeDocument points to ppt/presentation.xml. From there, each slide, slide master, and slide layout is referenced by its own relationship ID — the XML never uses raw file paths. This indirection means parts can be renamed, moved, or supplemented without breaking references, and tools can process the presentation graph without knowledge of the physical directory layout.

English Metric Units (EMU)

All spatial measurements in PPTX — slide dimensions, shape positions, font sizes in certain contexts — use English Metric Units. One inch equals 914400 EMU. One point equals 12700 EMU. One centimeter equals 360000 EMU. The value 914400 was chosen because it divides evenly by an unusually large number of common factors: 72 (points per inch), 96 (Windows DPI), 150 and 300 (common print DPI values), and 254 (centimeters × 100 per inch). Being an integer, EMU values serialize identically across platforms and languages with no floating-point ambiguity.

The default 4:3 slide size in presentation.xml reads <p:sldSz cx="9144000" cy="6858000"/> (10 × 7.5 inches). The default 16:9 widescreen size, standard since PowerPoint 2013, is cx="12192000" cy="6858000" (13.33 × 7.5 inches).

Slide master inheritance

PPTX implements a three-tier design cascade: slide master → slide layout → slide. The slide master (ppt/slideMasters/slideMaster1.xml) defines global placeholder positions, background fill, and default text styles. Slide layouts (ppt/slideLayouts/) customize these defaults for specific purposes — title slide, section header, two-column, blank, and so on. A default python-pptx presentation ships 11 layout variants. Each individual slide inherits from its assigned layout, and can override any inherited property at the shape level. This cascade minimizes redundancy: changing a font in the slide master propagates to every slide and layout that has not explicitly overridden that property.

Strict vs. Transitional conformance

ISO/IEC 29500 defines two conformance classes for OOXML files. Transitional (the default produced by Office 2007 through Office 2019) permits legacy features including VML (Vector Markup Language) graphics carried over from older Office versions. Strict (available since Office 2013) prohibits VML entirely, requiring DrawingML for all graphics. The Library of Congress maintains separate Format Description Documents for each: fdd000399 (PPTX Transitional) and fdd000402 (PPTX Strict). In practice, nearly all PPTX files in circulation use Transitional conformance.

Programmatic access

python-pptx (current version 1.0.2) provides full read-write access to PPTX files from Python. A script can create slides, insert text and images, build charts and tables, and read existing presentation metadata — all without any Office installation. Apache POI’s XSLF module offers equivalent capabilities in Java. Microsoft’s Open XML SDK targets .NET. These libraries work directly with the underlying XML, making automated report generation — populating slide templates with database data, generating weekly dashboards, batch-processing conference slides — a standard workflow in enterprise environments.

Security boundary: PPTX vs. PPTM

Standard .pptx files cannot contain VBA macros. Presentations with macros must use the .pptm extension, which carries a different MIME type (application/vnd.ms-powerpoint.presentation.macroEnabled.12). This separation — absent in the legacy PPT format — allows email gateways, antivirus scanners, and browser download policies to block macro-enabled presentations while permitting standard PPTX files. Embedded OLE objects and ActiveX controls remain a risk vector even in non-macro PPTX files, and external media links in relationship files can trigger network requests to attacker-controlled servers when the file is opened.