Audiobook Polish: clean ebooks to natural audio

Strips PDF junk and front matter, then turns ebooks into natural-sounding audiobooks

Problem

Most ebook-to-speech tools rely on engines like KokoroTTS and parse PDFs unreliably, reading front matter, table of contents, and OCR artifacts in a robotic voice; users want a clean structure and narration that sounds human.

Solution

After uploading a PDF/EPUB, use an LLM to identify and strip front matter, TOC, notes, and OCR noise, split by chapter, and clean the text; then generate audio with a more natural multilingual TTS model with pauses and intonation, exporting files ready for podcast or audiobook players.

Users

Readers, researchers, and content creators who want to turn their own ebooks, papers, or long documents into audio

MVP

Drag-and-drop PDF/EPUB upload; LLM cleaning and chapter splitting; pick voice and speed to generate audio; export mp3/m4b with chapter markers; charge by characters or minutes

Revenue

Usage-based billing by characters/minutes or a monthly subscription; batch conversion and API for content platforms

Why now

LLM text cleanup and high-quality TTS costs are both falling, so a solo dev can combine APIs into a polished experience while existing tools remain crude readers