Unified autonomous pipeline for planning, writing, and illustrating complete novels using MiniMax M3.
autonovel-minimax is an open source pipeline refactored by Alex Santos. Based on NousResearch's autonovel, this version consolidates fragmented API dependencies into a single provider (MiniMax M3), generating long-form prose and matching chapter illustrations under a unified key.
Concept generation, character worldbible, chapter draft synthesis, and PDF typesetting.
Design choices, architectural patterns, and engineering decisions implemented for extreme reliability and developer experience.
Eliminates API synchronization friction by using a single provider for both narrative prose generation and illustration synthesis.
Streamlined repository architecture with minimax_client.py and minimax_image.py for simplified maintenance and execution.
Injects previous chapter summaries and character knowledge states into each generation step to maintain continuity.
Builds formatted PDF and EPUB files featuring title pages, table of contents, chapter art, and balanced typography.
Real-world usage, terminal configuration, and technical integration commands.
# Run the autonomous novel generation pipeline
python main.py --premise "A cyberpunk mystery set in a flooded metropolis" --chapters 10 --illustrations --output-format pdfKey clarifications regarding licensing, compatibility, deployment, and security.
Typically 15 to 30 minutes depending on chapter count and image generation.
Yes, the pipeline pauses after stage 2 to allow manual outline refinement.
English, Portuguese, Spanish, Chinese, and many other languages supported by MiniMax.
Inside dedicated subdirectories under output/ with raw markdown and high-res images.