Turn handwritten Kindle Scribe notebooks into Obsidian Markdown.
OCR your handwriting — locally or in the cloud — review it side-by-side with the original, and export into your vault keeping your folder structure. Open source, runs on your machine.
What it does
Local or cloud OCR
Qwen3-VL via Ollama or LM Studio, fully private — or Claude/Gemini in the cloud for the hard pages.
Side-by-side review
Your original page next to the transcription. Fix the OCR before anything touches your vault.
Folder-mirroring export
Your Scribe folders become folders and wikilinks in Obsidian.
Real Markdown + math
Headings, tables, callouts, and handwritten math rendered as KaTeX.
Why not the Scribe's built-in convert?
Amazon's "Convert to text" emails you a flat .txt. marginalia drops it straight into Obsidian — folder structure, wikilinks, real Markdown and math — and lets you fix the OCR first. Fully local if you want.
How it works
Import
Drop a Scribe PDF, or point marginalia at your sync folder.
Review
Read each page next to the original and fix any OCR slips.
Export
Write Markdown into your Obsidian vault, structure intact.
Get started
One command installs everything and launches the app on your machine.
curl -fsSL https://raw.githubusercontent.com/VforVitorio/marginalia/main/scripts/install.sh | bash
irm https://raw.githubusercontent.com/VforVitorio/marginalia/main/scripts/install.ps1 | iex
The installer sets up uv + Python and downloads the prebuilt app — no Node needed. Local OCR also needs Ollama or LM Studio; cloud OCR works with a Claude or Gemini account.
FAQ
Is it free?
Yes — open source under the MIT license, and free to use.
Does my data stay private?
With local OCR, nothing leaves your machine. Cloud OCR (Claude or Gemini) is optional, for the pages a local model struggles with.
What if the OCR makes mistakes?
That's what the review step is for: you see the original page next to the text and fix any slips before anything is written to your vault.
Which devices does it support?
The Kindle Scribe only, for now. Other devices already have good tools (e.g. Scrybble for reMarkable).
For developers
Open source (MIT). Python + FastAPI backend, React frontend, with clean module boundaries and a one-command run. Contributions welcome — Scribe-only in scope.
Meet the author
