How I used PyMuPDF and the Kokoro-82M model to create a high-quality, local text-to-speech engine for converting PDFs to audiobooks.