PDF to Markdown, without rewriting it
Upload a PDF and get its text back as Markdown — headings, lists and table rows kept where the source has them. Files with a text layer are converted directly, usually in a second or two.
Markdown is the format that survives: it opens in Obsidian, Notion, Bear, VS Code and any plain text editor, and it still reads fine as plain text. The hard part is getting it out of a PDF without losing the structure — or letting a model paraphrase what your source actually said.
This mode does not rewrite anything. If the PDF carries a text layer, the text is extracted and normalised: no model call, no generation wait. If a page is a scan, that page is rendered at about 144 dpi and transcribed by a vision model instead — roughly twenty seconds per page, and only the first four pages.
What the output looks like
Real output: one photographed lecture page in, 542 characters of Markdown out, 7.6 seconds. Unedited:
# Lecture 6 - Cellular respiration Three stages: glycolysis (cytosol), the citric acid cycle (mitochondrial matrix), oxidative phosphorylation (inner membrane). Glycolysis nets 2 ATP and 2 NADH per glucose. Each turn of the cycle yields 3 NADH, 1 FADH2, 1 GTP, 2 CO2. Complexes I, III and IV pump protons; complex II does not. ATP synthase uses about 4 protons per ATP. Theoretical maximum: 30-32 ATP per glucose.
Who it is for
- Getting a PDF into Obsidian or Notion without retyping it
- Keeping the section structure so the text still makes sense after pasting
- Turning a lecture handout into something you can search and edit
Not a good fit: Two-column papers where the reading order is genuinely ambiguous, and PDFs that are mostly equations — symbols come out as text, not as typeset math.
Three things that improve the result
- One chapter at a time. A 300-page book is mostly extraction time and no extra value.
- Check tables after pasting: rows usually survive, merged cells often do not.
- If the source is a scan, expect about twenty seconds per page and only the first four pages.
How it works
- Upload a PDF — or a photo of a page if that is what you have.
- Keep Markdown selected as the output type and press the button.
- Copy the text, or download it as a .md file and open it in your notes app.
Questions
Does it change my wording?
No. For PDFs with a text layer the text is extracted as-is, including heading lines, list markers and table rows. This mode has no summarising step.
What happens with a scanned PDF?
Pages without a text layer are rendered to images and transcribed by a vision model, because there is nothing to extract. Only the first four pages are read, and a note above the result tells you when that limit was hit.
Which formats can I upload?
PDF, DOCX, PPTX, XLSX, EPUB, HTML, CSV, Markdown, plain text and images. Legacy .doc and .xls need to be saved as .docx or .xlsx first.
Is there a size limit?
40 MB per file. Files above that are rejected before any parsing starts.