Walking Book — Quick Start

A friendly, experiential tool for book editors.

Edit and engage from anywhere, unplugged from your desk. Expand your creative boundaries on the go.

What you do in Walking Book

• Create a tape: Turn your manuscript into a “Walking Book tape” (a .walkingbook / .zip file with narrated audio + timestamps).
• Play + mark edits: Listen in the Reader and record voice notes tied to the exact line you were hearing.
• Export: Download a Word .docx with your edits as native Word comments, or export a comments .json to merge into an existing DOCX.

Quick start (first-time users)

1) Create a Walking Book tape (Google Colab)

Click Create in the app (or open the Colab notebook directly):

• Colab notebook link (used by the in-app Create button): https://colab.research.google.com/drive/1Rxgki8JbduqqbcX9k63Z7T4jdXDVIZng?usp=sharing

In the notebook you will:

• Upload your manuscript (or paste text, depending on the notebook step).
• Pick a voice (Walking Book offers 18 English voices; you can preview them in the app under Voices).
• Run the cells to generate audio + timestamps.
• Download the output tape: a .walkingbook file (it may download as .zip).

Important:

• Do not unzip the tape. The Reader expects a .walkingbook / .zip file exactly as downloaded.

2) Load your tape in the Reader

• Open the app and press Play.
• If this is your first time, you’ll see an upload screen. Click Upload & open reader and select your .walkingbook / .zip.

Once loaded, your tape is stored locally in your browser (IndexedDB) until you export or wipe it.

3) Listen + record edits

• Press Play to start narration.
• When you want to leave a note:
  • Desktop: click the pencil / Edit button to open the Voice Editor.
  • Mobile + headphones: press pause on your headphones; the app listens briefly for the word “edit”. Say “edit” to open the Voice Editor hands-free.
• In the Voice Editor:
  • choose an Edit Type (Line Edit / Section Edit / Dev Edit),
  • speak your note,
  • say “I’m done” (or stop recording), then Save.

Your edits appear in the Edit Log and are attached to the closest timestamped text segment.

4) Get your edits back into Word

Open the menu → Comment Studio.

You have two main paths:

• Download DOCX with comments (easy mode): generates a new .docx where your edits are already inserted as Word comments.
• Merge into an existing DOCX (format-preserving mode):
  1. Download Comments JSON from Comment Studio (or use the current session comments),
  2. Upload your manuscript .docx,
  3. Click Merge & Download DOCX.

Tip: before wiping data or switching tapes, download a Session backup (so you can restore later).

Offline mode (recommended for walking)

On the homepage, click Use Offline. This prepares the app for offline use and downloads the Moonshine speech model for on-device transcription.

• Offline transcription: when offline mode is ready, the Voice Editor transcribes edits on-device with Moonshine.
• Online transcription: if offline mode isn’t enabled/ready, the app uses your browser’s built-in speech recognition (behavior and network usage varies by browser).

What is a “Walking Book tape”?

A tape is just a zip file with a known structure. The Reader accepts either:

• .walkingbook (recommended), or
• .zip (the same thing, just a different extension)

Expected contents:

• metadata.json (title, author, etc.)
• manifest.json (chunk list + timestamps)
• audio files referenced by manifest.json (often under audio/...)
• optional: version_history/*.txt (original text for better Word-comment placement)
• optional: session.json and edits/edit_<id>.webm (when exporting a tape-with-session backup)

Walking Book is for you

• Complete control: No AI suggestions or algorithmic interference. Just you, your words, and the freedom to think differently.
• Neurodivergent editors & beyond: Especially helpful for people who think better while moving — but useful for anyone who wants to break free from desk-bound editing. You can also use Walking Book on desktop, with OpenDyslexic font support.
• Reduce cognitive load: Listen instead of re-reading the same paragraphs over and over. Don’t accidentally read what you meant to write.

Run locally (developers)

Prereqs: Node.js + npm.

npm install
npm run dev

Then open the URL printed by Next.js (usually http://localhost:3000).

Privacy + your work stays yours

• Everything local-first: tapes, audio, and sessions live in your browser storage unless you export them.
• No AI training: your writing isn’t used to train models by Walking Book.
• Open source: Apache 2.0 license.

Network notes (so “privacy” stays concrete):

• No manuscript uploads by the app: there is no backend to send your manuscript/tape/session to.
• External links/embeds: the in-app Create button opens Google Colab, and the optional Quick Start guide can load from Gamma.
• Speech recognition:
  • Moonshine (offline mode) runs on-device (the model assets are served from this app’s own /vendor/ path).
  • Web Speech (browser speech recognition) may be cloud-backed depending on your browser.

🤝 Be a Founding Human (The Vibe Check)

Walking Book is 100% local-first. I have zero app telemetry or analytics — I don't know if you're using it, what you're editing, or if the colors look right on your screen.

I am looking for "Human Telemetry." I want to hear about anything you experience — joy, friction, or unexpected uses.

How to help: send a quick "Vibe Check" email to kate at sixpenny.org (or email kate@sixpenny.org). Just tell me:

• The Setting: Where did you take your Walking Book?
• The Palette: Which Riso color combo felt most "cozy" for your brain?
• The Friction: Where did the tech get in the way of your walk?
• The Material: Did you use it for a novel, a poem, or a dense textbook?

🛠️ Behind the Build (v0.8)

This is the third iteration of the system architecture. I originally explored cloud-based AI for live reading, but as a privacy professional (currently studying for the CIPP/E), I made a deliberate pivot.

Professional editors and authors handle sensitive, unreleased intellectual property. I chose to rebuild for 100% Local Privacy so that no manuscript data ever touches a cloud for AI training.

• Transcription: Powered by Moonshine STT (on-device).
• Narration: Powered by Kokoro TTS.
• Design: Inspired by the tactile parameters of my ten year old OP-1 portable field synthesizer by Teenage Engineering and Penny Magazine. With OpenDyslexic support special thanks to abbiegonzalez.com.

Notes / contact

Created by Kate Thomas at Sixpenny & Co.

Email thoughts, issues, ideas to: kate at sixpenny.org