Supported formats
MindMax reads these in place, out of your folder. There is no conversion step and no second copy.
| Format | Reading | Annotating | Search lands on | What the AI can change or make |
|---|---|---|---|---|
| PDF (with text layer) | Full page rendering, plus a reflowed text view | Yes | The page | The PDF is never changed. Revising it puts the edited text into a Markdown copy beside it |
| PDF (scanned) | Images until recognised | After recognition | The page | The same, once recognised |
| EPUB | Reflowable text | Yes | The passage | Nothing in the book. Translation writes a Markdown file |
Kindle (.mobi, .azw3) | Converted to EPUB, then read as one | Yes | The passage | As EPUB |
PowerPoint (.pptx) | Laid out as slides, plus a text view | Yes | The slide | Nothing in the deck. Translation writes a new .pptx with the layout kept |
Word (.docx) | Yes | Yes | The location in the document | Revised in place, as Word tracked changes by default. A review exports as a copy with Word comments |
Excel (.xlsx) | Yes | Yes | The location in the sheet | Nothing in the workbook |
Images (.png, .jpg, .webp, …) | Viewer and contact sheet | Ask about a region | The description, once read | Nothing in the image. A boxed region can become a Markdown note beside it |
| Markdown | Edited directly | Yes | The line | Anything, each edit a tracked change you accept |
See reviewing and revising with AI for revising Word files and PDFs, and translating a document for decks.
Search reaches into all of them and takes you to the location rather than naming the file.
In every one of these readers you can also drag a box around part of a page — a chart, a table, a diagram — and ask about just that region. The text under the box travels with the question, so numbers come from the document rather than being read off the picture.
Notes on particular formats
Scanned PDFs have no text until you recognise them. That is a step you start deliberately — see recognising scanned PDFs.
DRM-protected EPUBs cannot be opened. MindMax reads ordinary EPUB files.
Kindle files (.mobi, .azw3) are converted into a proper EPUB the first time you open one,
and everything downstream — chapters, annotations, quotes, jumping back to the source — treats
the result as an ordinary book. The conversion reads the table of contents the book ships with,
so a book that never had one opens as a single long chapter. Files bought from the Kindle store
carry DRM and cannot be opened.
Slides are laid out as slides — shapes where the author put them, speaker notes included in the text — rather than flattened into a list of bullet points. Quotes from a deck cite the slide number. Decks built by generative tools sometimes use decorative shapes in place of layout, which can shift the reading order of a page; the text view is there when a page looks wrong.
Images have no text of their own. They become searchable once they have been read — which happens when you ask a question about one, or ask for it explicitly — and what enters the knowledge base is the model's description, marked as such rather than quoted as if it were text printed on the page. HEIC and TIFF are converted for display on macOS only.
Multi-column PDFs with interleaved figures are the hardest case for any extractor. The characters are usually right; the reading order sometimes is not. Quotes from those pages are worth a click through to the page.
Badly built EPUBs — broken chapter metadata, or the whole book in one file — degrade chapter-level navigation, because there is nothing left to navigate by.
Not supported
- Video and audio files
- Formats behind DRM (including Kindle store purchases)
.doc,.pptand.xls(the pre-2007 binary formats)