# Audiblez Audiblez is an open-source tool that converts **EPUB e-books into M4B audiobooks** using AI text-to-speech. You give it an EPUB, pick a voice and language, and it generates chapter-by-chapter audio combined into a single audiobook file. ## Key points - **TTS engine** — Kokoro-82M, a compact (82M-param) text-to-speech model with natural-sounding output, trained on <100h of audio (Apache-licensed). - **Pipeline** — integrates FFmpeg ([[FFmpeg]]) for audio processing and espeak-ng for text analysis; outputs per-chapter WAV merged into M4B. - **Languages** — English, Spanish, French, Hindi, Italian, Japanese, Portuguese, Mandarin. - **Interfaces** — CLI and a GUI (`audiblez-ui`); interactive chapter selection; variable speed (0.5–2.0x). - **Performance** — GPU/CUDA accelerated; ~5 min for a ~160k-char novel on a Colab T4, ~1h on an M2 MacBook CPU. - **License** — MIT. A self-hosted, local-first way to turn your e-book library into audiobooks without a subscription. ## References - Repository: https://github.com/santinic/audiblez ## Related - [[FFmpeg]]