Extract text from PowerPoint (.pptx): Outline View, python-pptx, speaker notes

To extract text from a PowerPoint file, start with Outline View or Save As Outline for a one-off export. For code, use python-pptx to read text shapes and tables. If the slide is an image or screenshot, use OCR or image-based extraction instead.

Most .pptx files are easy to extract from — text lives in shapes that any tool can read. The hard cases are slides built as images, screenshots of dashboards, or decks exported to PDF where the original .pptx is gone.

1. PowerPoint's outline view

View → Outline View shows every text box as plain text in slide order. Select all, copy, paste into your destination. This catches title and body text but misses content inside grouped shapes, SmartArt, and image-based text.

Faster variant: File → Save As → Outline (.rtf). You get a clean text file of every text element on every slide.

2. python-pptx for programmatic access

MIT-licensed, handles every text shape including those in groups and tables:

from pptx import Presentation prs = Presentation("deck.pptx") for i, slide in enumerate(prs.slides, 1): for shape in slide.shapes: if shape.has_text_frame: for para in shape.text_frame.paragraphs: print(i, para.text)

Add an extra branch for shape.has_table to walk table cells. For grouped shapes, recurse into shape.shapes when shape.shape_type is GROUP.

3. Slides as images

If the deck is a series of image-only slides (common for branded marketing decks and screenshots-of-dashboards decks), neither of the above works. Two options:

• Export to PDF, then OCR with ocrmypdf or run through a PDF text extractor.
• Export each slide as PNG (File → Export → PNG), then run them through ExtractFox's image data extractor with a prompt like "extract all visible text in reading order."

Online .pptx files

If the file is on SharePoint or Google Slides, both support exporting to PDF or .pptx for free. The Google Slides API also exposes presentation content directly via REST — useful for automated pipelines pulling from a shared Drive.

python-pptx: tables, groups, and speaker notes

A complete extractor that handles the cases Outline View misses:

from pptx import Presentation from pptx.enum.shapes import MSO_SHAPE_TYPE def shape_text(shape): if shape.has_text_frame: return "\n".join(p.text for p in shape.text_frame.paragraphs) if shape.has_table: return "\n".join(" | ".join(cell.text for cell in row.cells) for row in shape.table.rows) if shape.shape_type == MSO_SHAPE_TYPE.GROUP: return "\n".join(shape_text(s) for s in shape.shapes) return "" prs = Presentation("deck.pptx") for i, slide in enumerate(prs.slides, 1): print(f"--- Slide {i} ---") for shape in slide.shapes: t = shape_text(shape) if t.strip(): print(t) if slide.has_notes_slide: print("[Notes]", slide.notes_slide.notes_text_frame.text)

Speaker notes live in a separate notes_slide — easy to miss if you only iterate slide.shapes. Compliance and legal reviews often care about notes more than slide body text.

Legacy .ppt files

python-pptx does not read binary .ppt (pre-2007). Options: open in PowerPoint/LibreOffice and Save As .pptx, or use LibreOffice headless:

soffice --headless --convert-to pptx legacy.ppt

Then run python-pptx on the converted file. Batch-convert a folder before your extraction script runs.

SmartArt and charts

SmartArt text is inside grouped shapes — the recursive shape_text function above catches most of it. Chart titles and data labels are trickier: chart shapes expose chart.chart_title and series names via the chart API, but data values require chart.plots[0].series[0].values. For a full data export, consider exporting the embedded Excel workbook (chart.part.related_parts) rather than parsing the visual.

Choosing the right method

What's inside a .pptx file

A .pptx is a ZIP archive of XML parts. Slide text lives in ppt/slides/slideN.xml inside <a:t> runs. Speaker notes are in ppt/notesSlides/. Tables, groups, and SmartArt each have their own part files. python-pptx abstracts this; if you need a quick grep without Python, unzip the file and search the XML:

unzip -p deck.pptx ppt/slides/slide1.xml | grep -oP '(?<=<a:t>)[^<]+'

That path is brittle — XML entities, split runs, and phonetic hints break naive regex — but it confirms whether text is actually stored or baked into images before you debug python-pptx.

Exporting slides to structured JSON

For a search index or RAG pipeline, flat text per slide isn't enough — you want slide number, title, body, and notes as separate fields:

def slide_to_dict(slide, index): title = "" body = [] for shape in slide.shapes: if not shape.has_text_frame: continue t = shape.text_frame.text.strip() if shape == slide.shapes.title: title = t elif t: body.append(t) notes = "" if slide.has_notes_slide: notes = slide.notes_slide.notes_text_frame.text return {"slide": index, "title": title, "body": body, "notes": notes}

Serialize the list with json.dumps and you have a document ready for embedding or full-text search. Keep slide numbers — citations back to the deck matter in compliance reviews.

Troubleshooting missing text

• Text in a picture — no XML text runs exist. Export slide as PNG and OCR.
• Embedded video with burned-in captions — captions aren't in slide XML; transcribe the video separately.
• Password-protected deck — python-pptx raises PackageNotFoundError. Remove protection in PowerPoint first.
• Macros in .pptm — python-pptx reads slide text fine; macro code in vbaProject.bin is ignored.
• Right-to-left languages — text extracts correctly but paragraph order may differ from visual order; test with your target locale.