Audiobook Creator

Version 1.2.0 - Convert Markdown books into high-quality MP3 audiobooks with state-of-the-art TTS models and hardware acceleration.

Features

• Multi-Model Support:
  • Supertone Supertonic: Ultra-fast ONNX-based high-fidelity models.
  • Soprano-80M: Lightweight LLM-based TTS with Vocos decoder.
  • MiraTTS: High-quality 48kHz audio tokens with NCodec.
  • Kokoro: High-quality open-weight model for expressive speech.
  • Chatterbox: Paralinguistic support for natural sounding books.
• Hardware Acceleration: Automatic detection of NVIDIA GPUs (CUDA), Intel NPUs/GPUs (OpenVINO), and DirectML.
• Smart Parsing: Automatically handles Markdown structures, headers, and semantic pauses.
• Dynamic Pauses: Uses semantic similarity to inject natural breathing room between sections.
• One-Click Setup: Professional PowerShell script handles all dependencies, models, and FFmpeg automatically.
• Modern GUI: dark-mode interface with real-time processing logs.

Quick Start (Windows)

1. Clone/Download this repository.
2. Right-click run_audiobook_creator.ps1 and select "Run with PowerShell".
3. Wait for the automatic setup (Installs UV, FFmpeg, and Models).
4. The GUI launches automatically!

Hardware Acceleration

Audiobook Creator automatically optimizes itself for your hardware:

Usage

GUI

Simply select your input file, choose a model, and click "Start".

Command Line

# Basic usage
uv run audiobook -i book.md -m soprano

# Specify voice and hardware
uv run audiobook -i book.md -m supertonic -v M2 --device cuda

Supported Models

---

Technical Setup

Prerequisites

• Python 3.12+
• FFmpeg (Auto-installed by script)

Manual Installation

pip install uv
uv sync
uv pip install -e .
audiobook-gui

Troubleshooting

• Security Policy: If PowerShell scripts are blocked, run: Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
• Missing Models: Ensure you have ~2GB free space for one-time model downloads.
• FFmpeg Output: If MP3 conversion fails, ensure FFmpeg is in your system PATH.

License

This project is licensed under the AGPL-3.0 License - see the LICENSE file for details.

Credits

Built with ❤️ by Amit Kumar.

Contributions are welcome! Please open an issue or PR for new model integrations or hardware optimizations.