AIStoryWriter
Python script that chains several language models to generate novella or novel-length stories from a prompt file.
Run outline, chapter and revision steps across local Ollama, Google or OpenRouter models.
AI Story Generator (AIStoryWriter) is a Python project that produces novella- to novel-length stories from a text prompt. It suits technically comfortable writers and hobbyists who want to experiment with local or cloud models.
What it is
You run Write.py with a prompt file and it drives a multi-step pipeline: initial outline, chapter outlines and chapter writing, each of which can use a different model. Models are given in the format provider://model@host, with ollama, google and openrouter supported. It can download missing models through Ollama and translate generated stories.
Who it's for
It suits experimenters who want to compare model combinations for long-form fiction. It is a poor fit for writers who want a guided editor, since the README itself lists repetitive phrasing, chapter flow and pacing as open problems.
Key features
Per-stage model choice
Use a different model for outline, chapter outline and chapter writing.
Ollama integration
Missing models can be downloaded automatically through Ollama.
Cloud provider options
Google and OpenRouter models are supported alongside local ones.
Story translation
The README says generated stories can be translated into other languages.
Editable prompts
Change prompts in Writer/Prompts.py or add your own prompt files.
Hardware guide
A Docs/Models.md page recommends models for different GPUs.
Up and running in 3 steps
Our honest verdict
What's great
- Each pipeline stage can use a different model, so you can pair a large model for outlines with a cheaper one for chapters.
- Local running through Ollama keeps your drafts on your machine.
- The README is candid about weaknesses, so expectations are clear before you start.
- A model recommendation table maps GPU capability to suggested models.
Where it falls short
- It is a command-line script run with ./Write.py after cloning the repository; there is no installer or graphical interface.
- Local models need real GPU power, and the sample config references models like llama3:70b that are heavy to run.
- Cloud use needs API keys copied into a .env file, and cloud providers bill by usage.
- The README lists repetitive phrases, uneven chapter flow and pacing issues as areas still being improved, and the repo shows no tagged releases.
Pricing & license
Free under AGPL-3.0 with no paid plan shown. Costs come from your own GPU hardware or from any cloud model provider you connect.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Do I need a GPU?
For local Ollama models, yes in practice; the README links a model guide by GPU capability and also offers Google as a cloud option for people without powerful GPUs.
Which model providers work?
The README lists ollama, google and openrouter.
What licence is it under?
AGPL-3.0. The README notes that if you modify the code and offer it as a network service you must share your modified source.
