Lipat

Appearance

PPTX import and export ​

parsePptx(bytes,{engine}) accepts an unencrypted PPTX ZIP and returns a scene, diagnostics and source-part metadata. File access, network requests, text decoding and delivery are owned by the host. No external image/link target is fetched.

Supported content includes presentation relationships and slide order, standard text/rich runs, direct RGB/common scheme colors, basic geometry, raster/static SVG images, six native category chart kinds, simple table content, speaker notes, inherited placeholder geometry and inherited decorative shapes. Groups are flattened with warnings; complex transforms are approximate. The parser resolves paths relative to each source part and rejects package-root traversal, duplicate ZIP names, malformed XML and resource overages.

engine.importScene(imported.scene) returns an editable canvas document. This is a separate explicit step. The parser does not require authoring/layout and does not infer grids or semantic components.

writePptx(scene,{engine}) writes a new canonical ZIP with a presentation, relationships, slides, blank layout/master/theme, embedded or linked images, native charts with embedded XLSX data, SVG vector/PNG fallback pairs, notes and native entrance timing. Reveal steps become click groups; supported animation fields become native effects. Easing approximations and timing rounding produce diagnostics. See animations and image URLs. Reopen exported bytes through the parser, or consume them in another application. Independent python-pptx verification checks writer interoperability.

Resource limits ​

  • Input ZIP: 32 MiB.
  • Expanded package: 64 MiB; individual entries: 32 MiB.
  • Package entries: 10,000; XML parts: 8 MiB and 200,000 XML nodes.
  • DTD/entity expansion is disabled.
  • Authoring documents: 10,000 slides, 100,000 nodes, 64 nested levels; requests: 96 MiB.
  • Embedded image asset bytes: 64 MiB total.
  • Reveal step: 0–10000 per node; PPTX export: at most 100000 reveal steps across slides.

These are implementation bounds, not performance guarantees. Memory/runtime costs depend on content. Use workers for latency-sensitive hosts.

Successful parsing does not imply arbitrary-PowerPoint fidelity. Unsupported content is omitted or approximated with diagnostics for detected features. See the complete support matrix.