Ask HN: Locally Run Text to Speech+AI: ePub|PDF –> Audio Book

5 points by mrjay42 ↗ HN
I've been researching this for a while, without success. Maybe, as I am not entirely connected to the AI-trendy world there's a solution for this?

I am looking for a text to speech (TTS) tool/toolchain. That would ideally transform a document such as an EPUB into an audio book.

An adjacent request would be, to understand HOW to generate locally the equivalent of "NotebookLM Podcasts".

If you don't know what those are:

* go there https://notebooklm.google.com/

* create a new notebook

* upload a document, like a big complicated book

* press "generate" on the right part;

* wait a little

* boom, now you can listen to an audio podcast entirely based on the book you sent where two fictional hosts discuss the book in question in a relatively thorough manner.

---

So yeah, I wonder what are the available, Free Software/Open Source, tools or tool chains that we can use locally on a reasonably powerful hardware at home to create text to speech using AI.

My ultimate goal would be to be able to generate audio books from EPUB/PDFs.

9 comments

[ 4.4 ms ] story [ 71.2 ms ] thread
Just a couple days ago, I saw this project mentioned. Looks promising but I haven't had the time to try it yet so this isn't an endorsement.

https://claudio.uk/posts/audiblez-v4.html

Thanks, I'm gonna check it out right now.
Just coming back here for a little report. This tool "audiblez" works fine and out of the box.

Pros:

* works out of the box

* text to speech available in many languages!!

* Audio output in .wav -> easily convertible into something else

* automatically creates chapters based on the book

The limitations are: * the voices are very "AI"-style, so they sound emotionless, we're far from anything resembling any kind of acting;

* the conversion of a book into an audiobook might just crash if the script encounters a character that it doesn't know how to convert.

Yes I've also tried it and reached similar conclusions. Do update it to the latest version though, as I did report an issue with non ASCII characters and for me it got fixed in v0.4.7+