Local audiobook pipeline

Most read-aloud features today are a thin wrapper around a big-tech cloud voice. I wanted audiobooks from my ebook library without any of my reading leaving the house, so I built the pipeline myself.

The synthesis end is Kokoro, a neural TTS model served from my home server with an OpenAI-compatible API, running entirely on CPU. The converter takes an EPUB or PDF and produces a proper .m4b: chapters derived from the book's table of contents, embedded cover art, and a voice picked to suit the material.

Around the converter sits the automation that makes it a pipeline rather than a script: a queue directory watched by a systemd timer, a render worker on the server, and an import step that files finished books into Audiobookshelf, where they sync to my devices. Queue a book at night, wake up to an audiobook.

The same Kokoro server also powers read-aloud in my desktop ebook reader through a small custom voice plugin, replacing a feature that would otherwise have gone through Microsoft's cloud.

What it demonstrates

Practical ML serving on modest hardware, audio and metadata wrangling with FFmpeg, and unglamorous-but-honest systems glue: queues, timers, idempotent imports. Built around a conviction that convenience should not cost you your privacy.