DEV Community

Cover image for Hacking the Narrative: How I Used NotebookLM to Write a Novel Where the AI "Human-to-Machine Transcoding" Chapter by Chapter
Saltmu
Saltmu

Posted on

Hacking the Narrative: How I Used NotebookLM to Write a Novel Where the AI "Human-to-Machine Transcoding" Chapter by Chapter

#generativeai #writing #gemini #notebooklm

The Concept: "Textual Gradation"

As an engineer and an author, I asked myself a question:
"Can I simulate the Human-to-Machine Transcoding not just through plot, but through the texture of the text itself?"

Most people use AI to write "clean" text. I wanted the opposite. I wanted to see if I could use Google's NotebookLM and Gemini 1.5 Pro to create a story where the prose starts rich and emotional, but gradually degrades into mechanical, logical, and cold output as the protagonist mechanizes his own body.

This is the "Post-Mortem" of my experimental novel, Clockwork Orpheus (Japanese: 機巧のオルフェウス).

Try generating your own text using this Shared Notebook

https://notebooklm.google.com/notebook/b7334e11-5f4f-418f-b1d2-4c435bf99668

You only need to execute one command in the chat:
"Write Episode X (1-32) in English."

The Architecture

To achieve a consistent yet evolving narrative over 50,000 characters (approx. 12 hours of work), I didn't rely on a single long chat context. Instead, I treated the novel as a software project using NotebookLM as my RAG (Retrieval-Augmented Generation) engine.

The Stack

  • Engine: Gemini 1.5 Pro (via NotebookLM)
  • Context Window: 1M+ tokens (handling all setting docs)
  • Input: 5 separate "Source Files" acting as databases.

The Source Code (Context Data)

I uploaded the following 5 files to NotebookLM. Think of them as the database schema for the story.

  1. Writing Policy (The "Config" File): Crucial. This document contains strict style rules.
  2. Plot Structure: The skeleton of the story.
  3. Story Overview: World-building rules to prevent hallucinations.
  4. Character Sheets: Detailed profiles.
  5. Glossary: Tech terms and unique nouns.

The Hack: Dynamic "Writing Policy"

The core trick lies in the "Writing Policy". I didn't just tell the AI to "write a story." I defined specific "Sentiment Parameters" for each chapter.

# Writing Policy Example
# Note: For English generation, replacing this with "Use Latinate vocabulary" or "Use passive voice" is effective

Phase_1:
  target_chapters: [1, 2]
  mode: "Emotional Mode"
  rules:
    DO:
      - "Heavily use modifiers related to the five senses (smell, temperature, discomfort)."
      - "Actively use metaphors for psychological descriptions."
    DONT: []

Phase_3:
  target_chapters: [6, 7]
  mode: "Mechanical Mode"
  rules:
    DO:
      - "Increase the ratio of Kanji and Katakana characters."
      - "Use report-style phrasing."
      - "Assert only 'facts'."
    DONT:
      - "All metaphorical expressions (e.g., 'like a...', 'seems like...')."
      - "All speculative expressions (e.g., 'might be...')."
Enter fullscreen mode Exit fullscreen mode

For the early chapters, the policy instructed:

"Focus on sensory descriptions. Use metaphors related to heat, pain, and longing. Prioritize the protagonist's internal emotional monologue."

As the story progressed, I updated the instructions (or prompted NotebookLM to reference specific "Phases"):

"Reduce sensory adjectives by 50%. Focus on objective facts. Describe events with logical causality rather than emotional reaction."

By the final chapter, the instruction was:

"Eliminate all metaphors. Use short, clipped sentences. Output must be strictly observational, like a system log."

The Result

The AI successfully executed this "Textual Gradation."

  • Chapter 1: Full of "pain," "heat," and "love." The sentences were long and winding.

The moment of waking always felt like the gasping breath before drowning.
My lungs demanded oxygen, and my heart reluctantly resumed its pulse. That lingering discomfort against the physiological phenomenon was the only proof I was still alive.
2046, Tokyo. The morning sun filtering through the gap in the blackout curtains looked dirty, like stage lighting meant only to illuminate dust. I reached for the left side of the bed.

  • Chapter 10: The prose became dry. The protagonist stopped "feeling" pain and started "detecting" damage.

Mid-layer area "Labyrinth Engine District." Time elapsed since entry: 14 hours. Rest: None. Complex three-dimensional structures and intermittent combat are draining resources.
Remaining ammo: 12 rounds average per person. Food: 2 solid bars. Water: 300ml remaining. Depletion is imminent.

  • Finale: A chillingly efficient, mechanical text that perfectly

mirrored the protagonist's complete loss of humanity.
Layer transfer: Complete. Environmental data: Updated.
Visual information: White. Organic textures completely deleted.
Walls, floor, ceiling—all constituted by white luminescent bodies.

Why This Matters for Developers

This experiment proves that "Style" is just another parameter.

We often treat LLMs as black boxes that output "average" text. But by architecting your context (RAG) and treating your prompts as dynamic configuration files, you can control the "Aesthetics" of the output with engineering precision.

I have released the full source (Japanese) on my note, including the NotebookLM link where you can test the generation yourself.

Note on the Source Material: The original setting documents are kept in Japanese. This is because the story involves complex metaphysical metaphors and cultural nuances related to the "Orpheus" myth that are difficult to translate directly without losing their intended literary context.

If you are interested in "Literary Engineering," give this method a try.

Top comments (0)