ppt2pptx

ppt2pptx converts Microsoft PowerPoint 97–2003 binary .ppt presentations to Office Open XML .pptx using only the Python standard library. It reads the CFB/OLE container and PowerPoint record stream directly; it never launches PowerPoint, LibreOffice, COM, or a conversion service.

The project is useful for a broad set of real presentations, but it is not a complete implementation of every legacy PowerPoint feature. Unsupported or lossy content is diagnosed instead of being silently presented as fully faithful.

Highlights

• Opens RC4 CryptoAPI password-protected presentations when a password is supplied.
• Preserves normal slide order and dimensions without mistaking masters or notes for slides, including append-only incremental saves with older document containers still present in the record stream.
• Resolves each slide's masterIdRef through the current master persist list, so layout/title-master backgrounds and decorations are taken from the referenced master rather than the first master in the file.
• Preserves hidden-slide state so excluded slides remain hidden during slide shows.
• Respects each slide's fMasterObjects setting when flattening master decorations, including PowerPoint's "Hide Background Graphics" behavior.
• Keeps flattened master shapes below slide-level shapes in the DrawingML stacking order, so master gridlines and decorations do not cut through editable foreground diagrams.
• Inherits vertical text anchoring from corresponding master placeholders when the slide placeholder does not store its own OfficeArt anchorText.
• Inherits master date/footer/slide-number placeholder geometry, alignment, font, color, and vertical anchoring for editable fields, while honoring fMasterObjects and DocumentAtom.fOmitTitlePlace visibility.
• Restores positioned editable text, fonts, sizes, colors, bold/italic/underline, paragraph alignment, bullets (including legacy Wingdings/Webdings glyph fonts, colors, and relative sizes), rotation, flips, and safe external hyperlinks.
• Preserves PNG, JPEG, GIF, TIFF, EMF, WMF, and PICT media, including cropping, position, rotation, and flips when present.
• Preserves bounded, validated embedded MSGraph.Chart.8, Equation.DSMT4, and package-backed Excel.Sheet.12 storages as editable OLE chart/equation/worksheet objects with their original vector previews; malformed, linked, and other unsupported OLE objects remain inert and precisely diagnosed.
• Recreates common editable shapes, including zero-axis horizontal/vertical lines without introducing a diagonal drift and legacy pct30 patterned line fills; preserves legacy arcs as native editable DrawingML presets instead of partial custom Bézier paths; plus solid/gradient/picture backgrounds, comments, speaker notes, slide numbers, dates, headers, and footers.
• Reconstructs legacy native tables (regular grids of rectangle cells) as editable DrawingML <a:tbl> graphic frames with preserved cell fills and 1-pt borders, instead of flattening them into scattered shapes.
• Copies common core properties such as title, author, keywords, and timestamps.
• Supports atomic single-file and recursive batch conversion with JSON diagnostics.

Installation

python -m pip install ppt2pptx

Command line

ppt2pptx presentation.ppt
ppt2pptx presentation.ppt -o presentation.pptx --report report.json
ppt2pptx protected.ppt --password-file password.txt
ppt2pptx inspect presentation.ppt --json
ppt2pptx batch input-directory -o output-directory --recursive --report batch.json

The input is always read-only and output files are written atomically. The converter refuses to overwrite its input.

Python API

The Python API accepts the password directly:

from ppt2pptx import convert

result = convert("protected.ppt", "protected.pptx", password="secret")
print(result.report.to_dict())

Current limitations

Charts, SmartArt, animation timelines, audio/video playback, embedded OLE objects, and complex freeform or grouped master geometry remain incomplete. Solid, common gradient, and stretched picture backgrounds are retained, while advanced fills and effects may be approximated. PICT data is preserved, but rendering depends on the PPTX consumer. PowerPoint 95 and earlier files use a different record format and are deliberately rejected with a clear error.

Unsupported or approximated advanced objects are reported with object-backed warning codes such as ANIMATION_OMITTED, AUDIO_OMITTED, VIDEO_OMITTED, MEDIA_ACTION_OMITTED, EMBEDDED_OLE_OMITTED, LINKED_OLE_OMITTED, ACTIVEX_CONTROL_OMITTED, CHART_OMITTED, DIAGRAM_OR_SMARTART_OMITTED, and COMPLEX_FREEFORM_OMITTED. Each warning includes count, record_types, and locations (slide_index, record_offset, object_kind) when a matching object is found. OLE diagnostics count external objects rather than their container/storage records and use ExObjRefAtom to identify the slide that owns each object. Legacy MS Graph and Excel chart ProgIDs inside an OLE container are classified as one slide-bound CHART_OMITTED object instead of also producing a generic OLE warning. Files with incremental-save history use only the latest ExternalOleObjectAtom per object ID and references inside current slide persist ranges, so stale chart or OLE revisions do not inflate warning counts. Files with those objects do not receive a blanket advanced-feature warning. A legacy AnimationInfo container and its child atom likewise count as one slide-bound animation object, not two record-level losses. PowerPoint 2002+ ___PPT10 timing trees are inspected for actual effect nodes and matched to the same legacy fallback by OfficeArt shape ID, avoiding both silent timeline loss and duplicate warnings for one effect. Text ParaBuildContainer settings are also bound to their timing effect by shape/build ID, so paragraph sequencing, direction, and automatic-delay loss is represented by that same object warning. Legacy chart build settings are linked the same way, while chart editability loss remains a separate CHART_OMITTED diagnostic. PowerPoint-saved SmartArt is detected from the owning shape's metroBlob DrawingML package and produces one slide-bound DIAGRAM_OR_SMARTART_OMITTED diagnostic when only its preview is retained. PowerPoint-saved media that reopens as a preview picture is detected through the owning shape's InteractiveInfoAtom II_MediaAction and produces one slide-bound MEDIA_ACTION_OMITTED diagnostic when playback behavior is lost.

Specifications

• MS-PPT: PowerPoint Binary File Format
• MS-CFB: Compound File Binary File Format
• MS-ODRAW: Office Drawing Binary File Format

Related projects

ppt2pptx follows the same package and command conventions as doc2docx and xls2xlsx:

Distribution	Python package / CLI	Conversion
msdoc2docx	doc2docx	DOC to DOCX
msxls2xlsx	xls2xlsx	XLS to XLSX/XLSM
ppt2pptx	ppt2pptx	PPT to PPTX

The projects share a consistent API, CLI structure, report format, and exit status conventions. Their conversion engines remain separate because the Word, Excel, and PowerPoint binary formats use different data models.

Development

python -m pip install -e .
PYTHONPATH=src python -m unittest discover -s tests -v
python -m build

The converter itself does not depend on LibreOffice or PowerPoint. The real-file regression script may use LibreOffice only to verify that generated packages render:

PYTHONPATH=src python scripts/validate_real_files.py tests/real_samples \
  -o tests/real_output --render --password hello

On Windows with Microsoft PowerPoint installed, bilateral visual regression exports per-slide PNGs from the source .ppt and converted .pptx, then writes metrics and diffs. Each side runs in an isolated DispatchEx instance; the report records the owned process IDs, and cleanup never targets unrelated PowerPoint processes. It also records PowerPoint's source/output object counts and per-slide structure differences for text, pictures, tables, charts, SmartArt, groups, OLE/media, comments, and speaker notes:

PYTHONPATH=src python scripts/make_visual_fixture.py -o tests/fixtures/visual_minimal.ppt
PYTHONPATH=src python scripts/compare_powerpoint_visual.py \
  tests/fixtures/visual_minimal.ppt -o tests/visual_evidence/visual_minimal

License

This project is licensed under the MIT License. See LICENSE for the full license text.