Convert PowerPoint to Markdown

Extract the substance from .pptx decks as structured Markdown — slide titles, bullet content, and tables — without opening PowerPoint. Entirely on your Mac.

How it works

Drop the .pptx into Milldown.Batch as many decks as you like — each converts independently.
Each slide's text content lands in order: titles, bullets, and tables.Layout and design are deliberately left behind — this is about the content.
Copy into your notes, your AI chat, or export retrieval-ready chunks.The Refine menu splits long decks into ~1k-token chunks with front matter.

What people use it for

Meeting prep — read a 60-slide deck as two pages of text instead of clicking through slides.
Knowledge bases — decks are where institutional knowledge goes to hide; Markdown makes it searchable in Obsidian or Notion.
AI summarization — paste the whole deck's substance into Claude as structured text it can actually follow.
RAG pipelines — chunked, front-mattered slide content indexes cleanly in a vector store.

A note on token costs. A .pptx really is mostly markup — 61,000 tokens of XML around 630 tokens of actual words. But that 99% figure would be measuring against something nobody does: no tool feeds a model raw slide XML, and the plain text extraction that does happen costs about the same as the Markdown. Milldown therefore shows no savings figure for PowerPoint.

The reason to convert a deck is that slide XML has no reading order a model can follow, while Markdown gives it one — slide titles as headings, bullets as lists, speaker notes as text.

Limits worth knowing

Charts, images, and SmartArt are visual objects — they don't have a faithful text form and aren't invented. What you get is every piece of real text in the deck, in slide order.

What comes across, slide by slide

Slide boundaries are marked, so a claim can be traced back to the slide it came from rather than becoming one undifferentiated wall of bullets.
Speaker notes are included, under their own ### Notes: heading beneath each slide. This surprises people: the notes are often where the actual argument lives, and a deck summarised without them is missing the reasoning.
Slide titles become headings, so the deck’s own structure survives into the Markdown outline.
Tables stay tables. A table drawn on a slide converts to a Markdown table, not to loose text.
Images become references with their alt text — whatever description the deck author set. Decks without alt text yield placeholder names like “Picture 4”, which tells you nothing; that is a property of the deck, not of the conversion.

Where it struggles

Layout and design are gone by design. Position, size, colour, animation and build order carry no meaning in Markdown. What survives is the content and its order.
A diagram made of shapes becomes its labels. Boxes and arrows drawn in PowerPoint convert to the text inside them, without the relationships between them — an architecture diagram arrives as a list of component names.
Text inside an image stays there. A screenshot pasted onto a slide is a picture; its words are not extracted unless you convert that image separately, where on-device OCR will read it.
Embedding images inflates token counts sharply. Keeping data URIs is off by default for that reason; leave it off unless the model genuinely needs to see the pictures.

Further reading: Measured token savings by file type · Preparing documents for RAG

Try it on your own files

Milldown converts PDFs, Office documents, web pages, and scans into clean Markdown — entirely on your computer. Free for 7 days.

Download for macOS

macOS 14+ · Apple Silicon · $39 one-time after trial (excl. VAT) · no subscription