My AI-Powered Audiobooks

Robots read my new audiobooks aloud, with feeling.

My first three books of essays are now available on Audible as audiobooks.

Labor Econ Versus the World: Essays on the World’s Greatest Market

How Evil Are Politicians: Essays on Demogoguery

Don’t Be a Feminist: Essays on Genuine Justice

Kudos to Anna Krupitsky for using this AI voice software to make the audiobook editions possible. I’m amazed by the quality. The AI even seems to “get” humor and irony.

Cool and eerie, all at once.

P.S. Further commentary on the production process from Anna:

Buying an audio book is still cheaper than creating one, so most people wouldn't want to create audio books just for themselves, unless they are huge fans and there is no readily available audio book. But it is still much cheaper than hiring an actor to do it, and costs less than equipment that allows you to read it yourself, as well as less trouble. Eleven labs charges $99 for about 1MB, so creating one audiobook, which is about 300-400GB would cost around $30-40 without editing (and the amount of additional editing depends on the preparation).

Then there is preparation and post production.

Preparation involves converting PDF to Word, then converting Word to text document. Removing references, footnotes, references to footnotes, and page numbers. Each paragraph has to be a line. Mark the beginning of new chapter, as it would give the AI a way to distinguish chapters and new paragraphs.

Then run a script that reads each line and sends it to Eleven Labs API. The script creates file for each line in the text document. For example, line 5 will create 005.m3 file. The script then merges these files into chapter mp3 files.  During the merge insert pauses (silent mp3 files) between paragraphs and sections. I didn't know this, but numbers need to be put into words, as it cannot read the digits, so it took me more additional editing.

Then run each chapter file though Audacity to convert them to 132KB mp3.