PDF to mind map, the real pipeline: parse → summarize → generate
Much of what you want to map lives in PDFs, Word files, or scans that cannot be pasted. The Document to Markdown tool now offers a "Parse and create mind map" one-click path that sends the parsed text to AI in parts (up to 60,000 characters, 8 parts); beyond those limits, the manual "parse → AI assistant → mind map" pipeline still works with every step reviewable.
Step 1: parse the document into Markdown
Open the Document to Markdown tool and pick a file (PDF, Word, PPT, Excel, or common images, up to 4 MiB), or paste a publicly accessible document URL (still subject to the upstream service’s size and quota limits). Enable Force OCR for scans.
Your file or URL goes to this site’s backend and is processed by the MinerU parsing service with queuing; the page shows queue and page progress, and returns clean Markdown.
- Open Toolbox → Document to Markdown
- Pick Upload file or Document URL; enable Force OCR if needed
- Click Parse and wait for progress to finish
- Copy or download the resulting Markdown
Step 2: compress it with the AI assistant
Parsed output often runs to tens of thousands of characters — beyond the 12,000-character generation limit. Click Send to AI assistant and run Summarize or Key points to compress it; oversized results are truncated to the first 12,000 characters, so long documents are best parsed and summarized in sections.
Step 3: generate and refine the map
Click Generate mind map on the assistant result — the text is handed to the workspace and a map is generated automatically. Refine with Expand or Simplify, then export SVG or PNG.
Example input → Example output
Example input
(excerpt sent to the AI assistant)
Chapter 3 Photosynthesis
3.1 Light reactions: on the thylakoid membrane, water splits into ATP and NADPH…
3.2 Calvin cycle: fixes CO2 in the stroma…Example output
# Photosynthesis
## Light reactions
- Site: thylakoid membrane
- Output: ATP, NADPH, O2
## Calvin cycle
- Site: stroma
- Process: fixes CO2 into sugar
## Relationship
- Light reactions supply energy; the cycle fixes carbonLimitations
- Files are limited to 4 MiB; split bigger documents — the public-URL option is bound by the same upstream limits and does not accept unlimited sizes.
- Daily MinerU quota is limited — expect queuing or a quota error at peak times.
- Scans depend on OCR; blurry pages may produce wrong text.
- Files and URLs leave your browser for the parsing service — do not upload confidential documents.
- AI summaries can drop details; verify quotes against the original.
FAQ
Can I upload a PDF and get a map in one step?
Yes — use "Parse and create mind map": the file goes to MinerU for parsing, then the text is sent to the AI model in parts and merged into one map. Documents over 60,000 characters or needing more than 8 parts still require the manual summarize-then-generate pipeline. Retry covers only missing parts and does not re-parse.
How do I handle a hundred-page report?
Parse within the 4 MiB limit, summarize section by section with the AI assistant, then merge into one map. Dumping the full text in one go exceeds the input limit.
Is my uploaded file stored?
The file is forwarded through this site’s backend to the MinerU parsing service and the Markdown is returned; whether the third-party service retains anything is governed by MinerU’s own policies, which this site cannot guarantee. Do not upload sensitive documents.
Can scanned image-only PDFs be read?
Yes — enable Force OCR. Quality depends on scan clarity, so review the Markdown before continuing.
View template: Reading notes framework