SYSTEM NOTICE

Auto translation by AI. Be sure, accuracy, nuances and authorial intent may not be fully reflected.
見出し画像

Become an AI Editor! 5 Steps to Create a Book Using Your Past Note Articles [KDP Debut!]

This article is a continuation of the previous one where I highly praised the note I created.



✅ General Flow: 5 Steps to Becoming an AI Editor


🧭 How to become an AI editor?

Instead of "leaving everything to AI," by
combining human editorial skills with AI's structural capabilities, you can automatically generate "text that readers want to read."


🛠 Workflow Overview (5 Steps)

  1. Export note articles (bulk acquisition in XML format)

  2. Chunk articles into individual pieces using Python (format titles and body text)

  3. Select articles and combine materials (compose one chapter)

  4. Generate chapter structure with Claude or ChatGPT (API or Web)

  5. Save/organize in .docx, etc., for publication (supports Word, PDF, EPUB)


✅ Detailed Explanation of Each Step


① Export articles from note.com

  • Download all articles in bulk in XML format from note's "Settings → Export Function"

  • It is in the same RSS format (WXR format) as WordPress, including images and drafts



② Chunk and format articles with Python

  • Read the XML and split it into one article per <item>

  • Extract the title and body text, and remove tags (remove <br>, <img>, etc.)

  • Include "title + character count" in the filename to improve manageability

  • Result: Materials like article_chunks_cleaned/001_TitleName_2450characters.txt are completed

[Title] Double Your Concentration! 🌟 Pomodoro Technique BGM Created with Blue and Yellow ⏲️
[Character Count] 2517 characters
What is the Pomodoro Technique? ...

It's something like this.


3. Select only the articles you want to use and combine them (create materials for Chapter 1)


  • Automatic combination via script -> Output as chapter_01_merged_13429characters.txt

  • Optimized as material to feed into Claude or GPT

Concrete example)
First, prepare the articles divided as shown in the figure below.

Next, select the articles to be included from among them. In the case of the figure below,
put them in a folder called in_merge, then use Python to put the combined text data into a folder called out_merge and feed that to the AI.


4. Automatically generate chapter structure manuscripts with Claude or ChatGPT

  • Structure the text using Claude Web's Explanatory mode or the Claude API

  • In the prompt:

    1. ・Please make it 5,000 to 6,000 characters ・Please include chapter titles and subheadings ・Please include an introduction and a summary

  • Save the output text as .docx


5. Save and organize the manuscript -> Book publication (KDP or Note serialization)

  • If you convert .docx to .md or .epub, you can also support Kindle publishing

  • Serializing on Note as an "AI-edited reconstructed version" is also fine


✅ Mindset required for an AI editor

  • ❌ It is still too early to say "I want the AI to do everything"

  • ✅ "Select and prepare materials yourself" -> "Leave the structure and formatting to the AI"
    This is currently the most realistic and high-quality approach

I wonder what will happen eventually...


✅ Tools and Libraries Used (All free or pay-as-you-go API)

  • Python (Article formatting / Material preparation)

  • BeautifulSoup (HTML removal)

  • Claude Web or API (Generation)

  • python-docx (Word output)

  • note XML export feature (Material acquisition)

That's probably about it.

※Please note that the Claude API can be very expensive!


✅ Results of the practice

  • Claude Web: Stable generation of manuscripts around 4,000 to 5,000 characters

  • Claude API: Opus can output up to 4,096 tokens, and Sonnet can output up to 8,000 tokens

  • Compared to GPT-4 Turbo, Claude's writing style is 'softer and has a more natural structure'

  • Fully automated from material organization to output for one chapter (reproducible structure examples included)

This is only what is possible as of 2025/04/09!
AI evolves very quickly, so please do not take the above content at face value.


✅ Conclusion: Why is the 'AI Editor' valuable now?

There are a huge number of high-quality articles buried in notes and blogs.

By not just collecting them, but systematically organizing them into a 'book structure,' the reader's experience changes dramatically.

However, doing such editorial work alone requires a great deal of effort. In the coming era, the role of an 'AI Editor' who works in cooperation with AI to advance editorial tasks will become important.
AI Editor
will become important.



いいなと思ったら応援しよう!