Audiblez is a tool that generates audiobooks in m4b format from EPUB e-books using text-to-speech synthesis.
The tool addresses the challenge of converting written books into listenable audio by leveraging Kokoro-82M, a lightweight text-to-speech model with 82 million parameters that produces natural-sounding speech. Audiblez processes EPUB files chapter by chapter, generating individual WAV files and then combining them into a single m4b audiobook file. The approach supports multiple languages and offers adjustable playback speed between 0.5x and 2.0x. Users can select from various voices, each identified by language code and speaker gender.
The tool suits anyone wanting to convert their personal EPUB library into audiobooks without relying on commercial services. It works via both command-line and graphical interface, making it accessible to users with varying technical comfort levels. Installation requires Python 3 along with espeak-ng and ffmpeg dependencies. Performance varies significantly by hardware: on a GPU like Google Colab's T4 via CUDA, a typical novel converts in around five minutes, while CPU conversion on consumer hardware takes approximately an hour. Windows users should install the tool in a Python virtual environment, and CUDA support requires separate PyTorch installation. The graphical interface requires additional dependencies beyond the core command-line version.
The project maintains active development with recent additions of GPU acceleration and a graphical interface. The codebase shows ongoing refinement of installation procedures, particularly for Windows compatibility. Documentation includes specific performance benchmarks and detailed setup instructions for different platforms and hardware configurations.
