How to make an audiobook with an AI voice
An AI narrator lets you turn a manuscript into an audiobook without a studio: one consistent voice for the whole book, every chapter redone in minutes when the text changes. The hard part is not the voice but the workflow: preparing the text, keeping chapters consistent and exporting files a store or a player accepts.
This guide walks through that workflow with the SuperClonica catalog and the free SuperClonica engine, which voices long texts offline on an ordinary computer.
Choose a narrator you can listen to for hours
Long listening needs a calm, even voice: a narrator or a conversational male or female voice rather than a character. Try a full paragraph of your book on the page of each candidate (30 to 100 characters per test) and pick the one that still sounds pleasant on the tenth line.
Prepare the text
Clean text matters more for an audiobook than for anything else: the listener cannot skip back to see what a word was.
- Split the book into one plain .txt file per chapter.
- Remove page numbers, headers, footnote marks and hyphenation from the layout.
- Write unusual names, numbers and abbreviations the way they should be read, and keep a list so every chapter uses the same spelling.
- Use punctuation for rhythm: paragraph breaks and full stops give the narrator room to breathe.
Voice the chapters offline
Buy the voice once, download its file and install the free SuperClonica engine: python -m pip install superclonica-cpu. Then voice each chapter with one command, for example superclonica --text "$(cat chapter01.txt)" --voice narrator.json --lang en --output chapter01.wav. The engine splits long texts at natural pauses and joins them into one file per chapter; every chapter starts from the same timbre, so the narrator sounds the same throughout the book.
By our estimate the engine voices speech 6.4 to 9.7 times faster than real time on a desktop 8-core processor, so an hour of audio takes roughly 6 to 10 minutes, without a graphics card.
Listen, fix, re-voice
Listen to each chapter at normal speed with the text in front of you. When a word is misread, change its spelling in the text and re-voice that chapter only: it takes minutes, not a new studio session.
Export for players and stores
The engine writes WAV files. Convert them with any audio tool, for example ffmpeg, to MP3 or to an M4B audiobook with chapters, and normalize the loudness the store asks for. Before publishing, check the store’s rules about AI narration and make sure you have the rights to the text and the voice.