Notebook to JSON,structured and scriptable.

Use the converter at the top of this page: choose or drop in your .ipynb file and the browser writes a normalised JSON document — summary first, then every cell in notebook order. Download the .json file. No upload, no install, no terminal.

Because the raw file is not a document you can rely on. Cell sources are sometimes strings, sometimes arrays of lines; outputs mix mime bundles with base64 payloads; metadata differs between tools. This converter pins one shape for all of it and adds a summary, so a script written once keeps working.

With jq '.summary' notebook.json you get the cell counts, language and output totals in one line; jq '.cells[] | select(.cell_type=="code")' walks only the code. The summary object exists exactly so these one-liners do not need to understand nbformat.

total_cells, code_cells, markdown_cells, raw_cells, outputs, cells_with_image_outputs and code_lines — plus the language, kernel and nbformat at the top level of the document. Everything needed to decide whether a notebook is worth parsing further.

Yes. Base64 plots and figures are usually 90% of a notebook's bytes. With image stripping on, each payload is replaced by its mime type listed in stripped_mimes, so you can see exactly what was left out instead of losing it silently.

Yes — outputs and execution_count are kept per code cell by default. Stream text, results and tracebacks all carry over, with ANSI colours removed so the text is clean inside a JSON string.

Run jupyter nbconvert --to json notebook.ipynb. It writes the notebook back out as-is: the same structure, the same base64 payloads, no normalisation and no summary. Use it when you need the untouched file; use this page when you need a clean document.

Images attached to markdown cells are rewritten as data URLs in the JSON, so the document stays self-contained — one file with no sidecar images — and a renderer reading the JSON can still resolve them.

Yes. The same notebook serialises to the same bytes in a fixed key order, so diffs show real changes only. Combined with stripped images, it is the compact, diffable form of a notebook for repository review.

Yes. The file is read with the browser File API and serialized inside the tab, so after the page loads you can disconnect the network and the conversion still works — which is also the practical proof that nothing was uploaded.

No. The file is parsed as JSON and re-serialized inside the tab. No server is involved at any step, and closing the tab leaves nothing behind.

Yes, with no account, no file-size gate and no watermark. The conversion runs on your own device, so there is no per-file cost — same as the site's IPYNB to PDF, PY, HTML, Markdown, LaTeX and DOCX converters.