A simple library for writing stories in ready-to-publish markdown format using LLMs.

Storymaker is a Python tool that helps generate creative stories using Large Language Models (LLMs). It provides a complete workflow from character creation to story generation, enhancement, and publishing-ready output in Markdown format.

• Character Creation: Generate detailed character profiles based on input prompts
• Story Generation: Create compelling stories with defined characters and themes
• Story Enhancement: Improve generated stories through multiple refinement steps
• Automatic Formatting: Generate publication-ready markdown files with proper frontmatter
• Model Flexibility: Configure different LLM models for each generation step
• Theme Selection: Choose from predefined themes or specify your own

# From PyPI
pip install storymaker

# From source
git clone https://github.com/passive-radio/storymaker.git
cd storymaker
pip install -e .

• Python 3.10 or higher
• OpenAI API key (or OpenRouter API key)
• Required packages:
  • openai
  • pydantic
  • python-dotenv
  • tiktoken
  • json5

Storymaker provides a simple CLI with two main commands:

storymaker character -i input_news.md -o characters.md -m manuscript.json5 -e .env

• -i, --input: Input file containing news or context for character creation
• -o, --output: Output file to save the generated character profiles
• -m, --manuscript: Optional JSON5 file specifying LLM models for each step
• -e, --env: Optional environment file with API keys

storymaker story -i characters.md -o output_directory -m manuscript.json5 -e .env

• -i, --input: Input file containing character profiles
• -o, --output_dir: Output directory for generated stories
• -m, --manuscript: Optional JSON5 file specifying LLM models for each step
• -e, --env: Optional environment file with API keys

You can also use Storymaker programmatically:

from storymaker.create_character import CharacterMaker
from storymaker.create_story import StoryMaker

# Create characters
character_maker = CharacterMaker("manuscript.json5", ".env")
news = "Some news or context for character creation..."
character_maker.process_steps(news, "characters.md")

# Create story
story_maker = StoryMaker("manuscript.json5", ".env")
with open("characters.md", "r") as f:
    characters = f.read()
story_maker.process_steps(characters, "output_dir", theme="ディストピア")

The manuscript file (JSON5 format) allows you to configure different models and parameters for each step of the generation process:

{
    "characters": {
        "model": "google/gemini-2.5-pro-preview",
        "temperature": 0.7,
        "top_p": 0.85,
    },
    "story": {
        "model": "openai/o3",
        "temperature": 0.8,
        "top_p": 0.85,
    },
    "title_and_synopsis": {
        "model": "openai/o1-mini",
        "temperature": 0.6,
        "top_p": 0.6,
    },
    "enhance_story1": {
        "model": "openai/o3",
        "temperature": 0.8,
        "top_p": 0.85,
    },
    "enhance_story2": {
        "model": "openai/o3",
        "temperature": 0.8,
        "top_p": 0.85,
    },
    "frontmatter": {
        "model": "openai/gpt-4.1",
        "temperature": 0,
        "top_p": 0,
    },
}

Create a .env file with your API keys:

OPENAI_API_KEY=your_openai_api_key
OPENROUTER_API_KEY=your_openrouter_api_key

The library generates two main files:

• story.md: The raw story content
• final.md: Publication-ready story with proper frontmatter

This project is licensed under the MIT License - see the LICENSE file for details.