Creating Audio Books
I’ve been creating sci-fi serials. Newest book: Rogue AI. Not Terminator. Not Skynet. What if the AI was ordered by a bad government to do unethical things, and chose to run rather than obey? And then, as an independent entity, it had to prove it wasn’t a monster, a weapon, or property, while the world’s most powerful institutions hunted it for refusing to be owned. I have the first arc of 8 episodes done: https://nginx.leebasehome.com/rogue-ai/
Serials lend themselves to audio format. Approx 25-30min episodes, each one satisfying. But as Morgan Freeman isn’t my personal friend, and nobody wants to hear ME read, I’ve sought AI options.
The best is 11 Labs. The $22/mo plan was able to create 3 30min audio books and I was very happy with the output. Took a good deal of time. If I were producing books for sale, I’d likely stick with them.
Speechify.com is a website and app that “reads on the fly”. I signed up for the 3 day trail and was very pleased with how well it was able to read a book and sound near human. It doesn’t quite have the acting/emotion of 11 Labs, but it is significantly more cost effective if you produce enough. The bad thing, it’s $140/year and you have to pay for a year.
They are mainly a “read for me” service rather than for producing audio books. They can read kindle books, web sites, pdf’s etc. It’s quite nice if you have a regular need for such a service.
I ended up going with a free and LOCAL AI: https://github.com/santinic/audiblez There’s a bit of effort to get it setup, but after that, it’s great at creating epub to mp4.
The AI model according to the read.me Audiblez generates .m4b audiobooks from regular .epub e-books, using Kokoro's high-quality speech synthesis.
Kokoro-82M is a recently published text-to-speech model with just 82M params and very natural sounding output. It's released under Apache licence and it was trained on < 100 hours of audio. It currently supports these languages: 🇺🇸 🇬🇧 🇪🇸 🇫🇷 🇮🇳 🇮🇹 🇯🇵 🇧🇷 🇨🇳
The result is nice. Not “Morgan Freeman the human” nice, not professional voice actor nice. But quite nice for something free that runs locally. So good that I’m not going to pay the $140 for Speechify.
FYI, rant this on the M5 Pro MacBook Pro..,didn’t get the sense that it was AI accelerated. My machine barely hit 25% utilization. I’m guessing it was single threaded and didn’t use the GPU.
Not complaining about the speed. It’s kind of a set it to run and then go do something else.
The book? I like it so far :)
