Convert Word, Excel and PowerPoint to Markdown for AI
Turn .docx, .xlsx and .pptx files into clean Markdown offline, check the token count, and split long documents so they fit your AI model's context window.
Your progress
Convert Word, Excel and PowerPoint files to Markdown so an AI reads the content instead of the formatting: headings, lists and tables survive, slides stay in order, and spreadsheets become tables. 1FileTool does it offline, with no Microsoft Office installed, then tells you whether the result fits your model.
What you'll need
- 1FileTool 1.11.1 or later on macOS 13+, Windows or Linux.
- Your own
.docx,.xlsxor.pptxfiles. Older.doc,.pptand.rtffiles also work if LibreOffice is installed (see step 2). - On the free plan, conversions count toward the 8 files a day.
Protect your originals first
Replace source is ON by default: the output overwrites the original file and no backup is kept. Before following along, open Settings › General and turn Replace source off, or pick a separate output folder.
This one matters: with Replace source on, the Markdown is written over the original document, so report.docx would end up holding Markdown text. Turn it off for this run (or set the default in Settings › General) and the .md files go to your output folder while your documents stay as they are.
Steps
1. Open Office to Markdown
In the sidebar, open Office and pick Office to Markdown under For AI. For PDFs, HTML, EPUB, CSV, JSON or XML, use AI Context › Convert to Markdown instead; it takes those formats as well as Office files and can Read scanned pages with OCR. Home's Doc to Markdown quick action opens it too.
2. Drop the files
Drop one document or a whole batch onto the drop zone ("Word, Excel, PowerPoint — .docx, .xlsx and .pptx and the legacy formats"). Modern files are read directly. If a file needs LibreOffice and you do not have it, 1FileTool lists it as skipped before you run and offers Set up LibreOffice, which points Settings › Add-ons at a copy you installed yourself. 1FileTool never downloads it for you.
3. Set the two options that matter
- Speaker notes ("Include PowerPoint notes in the output.") is on by default. Turn it off when the notes are for the presenter, not the reader.
- Rows per sheet ("Cap on Excel rows, so one export cannot flood a prompt.") defaults to 2000. Lower it for big workbooks; the model rarely needs every row to answer a question about the data.
4. Run it
Press Run on 3 documents. Each file is converted on its own, so one damaged document does not stop the batch, and report.docx and report.xlsx in the same run both keep their output instead of one overwriting the other.
5. Copy it or hand it off
Copy Markdown copies the first result to your clipboard. Ask an agent opens the results in AI Chat with the assistant on your machine. Reveal in Finder (or File Explorer) shows the .md files.
6. Check the cost before you paste
Open AI Context › Token Counter, set Token model to match your AI (GPT-4o, GPT-4, Claude or Rough), and drop the Markdown files. The bars show which context windows each file fits and which it overflows.
7. Split anything that does not fit
If a document is too long, open AI Context › Split for Context, drop the file, and set Tokens per part (8000 by default) and Overlap (200 tokens repeated from the previous part). Splits prefer paragraph breaks, then line breaks, and the parts are numbered <name>.part1.md, <name>.part2.md and so on (part01 once there are ten or more).
The result
One clean .md file per document in your output folder: Word headings as Markdown headings and tables as Markdown tables, every sheet of a workbook under its own ## <sheet name> heading, and every slide in order as ## Slide 3 — <title> with its speaker notes quoted underneath. You know its token cost before it goes anywhere, and long documents arrive as parts that each fit your model.
If it didn't work
- "this format needs LibreOffice installed.": the file is
.doc,.ppt,.rtfor OpenDocument. Install LibreOffice, then point Settings › Add-ons at it, or save the file as.docx/.pptxfirst. - A Word file will not open after the run: Replace source was on, so the Markdown replaced the document's contents. Restore it from a backup or Time Machine / File History, and keep Replace source off for conversions.
- A spreadsheet came out cut short: that is Rows per sheet. Raise it and run again.
- A scanned PDF came out empty: use AI Context › Convert to Markdown with Read scanned pages on. OCR is much slower, and English only.
- The Claude count differs from what Claude reports: Claude's tokenizer is not public, so Claude (closest match) is an estimate. Leave headroom.
Try next
Need a whole folder of code or documents in one go? Pack a project folder into one file. Sending the original document outside your team instead? Remove the author metadata first.