Audible AI Narration

[ThisAdamGuy]

I just got an email from Amazon telling me I'd been selected to try out Audible's new feature of narrating audiobooks with AI. While I'm usually against using AI for creative projects (at least ones you intend to make money from) I admit I was curious. So I booted up the audio toolset and loaded Henry Rider and the NuYu Prescription into it. The result was immediately disappointing. The narrator sucks, first and foremost. You'd get about the same quality from putting your book into Google Translate and clicking the Read It Our Loud button. And there's something weirdly sad about hearing an emotionless robot reading Henry's highly emotional dialogue.

If they had an actual high quality AI doing the narration that actually sounded like it knew what it was reading, I might have been tempted to go through with it. Audiobooks are expensive to make, and it's probably the only way a lot of self-published authors will ever be able to get into that market. But the way it is now, you may as well get a Speak and Spell to narrate it for all the life these voices have behind them.

 

[Macha]

High quality.
AI.

Choose one.

 

[ThisAdamGuy]

High quality.
AI.

Choose one.

I'm not a snob, so I can admit that AI can churn out some good stuff. I'm just against that stuff being sold.

 

[3guanoff]

There are plenty of AI-narrated audiobooks being distributed right now. Some of them are AI-written as well. If I want a robo voice, I use TTS.
Take this gem for example:

The Greek Pantheon Audiobook on Libro.fm

Greek mythology is a vast and intricate tapestry of gods, heroes, and monsters that have shaped the cultural and literary traditions of the Western world. These myths were more than just entertaining stories; they served as explanations for natural phenomena, religious rituals, and human...

libro.fm

Straight out of ChatGPT, this audio-"book" is full of filler and fluff with little substance.
And if you doubt that "Betty Fisher" is an AI voice, listen to this treasure read by "Patricia Henders":

Roman Society and Culture Audiobook on Libro.fm

Roman society, one of the most influential civilizations in human history, was built upon a complex and well-structured foundation that evolved over centuries. The origins of Roman civilization can be traced back to the early settlements along the Tiber River, where communities of farmers and...

libro.fm

Those two must be twins separated at birth. Yanabo Nava LLC is so lucky to recruit both these talented narrators with identical voices.

Fraud at its finest.

 

[ArtBusterBeeze]

I think you'd better start taking any Ai offers with a grain of salt. There's no such thing as instant success with them, and these companies are trying to take something because it's 'hip' and run with it without putting in any real effort to make it good for the people. They are not Artists.

 

[LiteraryWho]

I'm not a snob, so I can admit that AI can churn out some good stuff. I'm just against that stuff being sold.

The DuoLingo voice model is unironically amazing, but I suspect it's a lot of work to make it sound as good as it does (in terms of telling the model when and how to emote and such). Given the lack of ready made training data (i.e. books with the sections already tagged), I suspect a general use AI Audiobook-inator (that doesn't suck and is easy to use) is a long way off.

 

[RepresentingCaution]

Thanks for the warning!

 

[ConansWitchBaby]

Is there an AI voice that does it in Jive? If so, I'm willing.

 

[CharlesEBrown]

If I understand their system correctly, I'm 2000 words short of finding out what PocketFM's "reader" (AI) sounds like on one of mine.

 

[John_Owl]

High quality.
AI.

Choose one.

I found one a while ago that was at least mid-quality. But with proper tweaking, the only thing missing was the audible breaths that you hear but don't really pay attention to from a human speaker. It could take direction for more emotional pinpointing, you could use different voices, so you'd have a narrator and a different voice for each character, and you could adjust speed and pitch for each section as well. The issue was that it had a 500 word trial and cost like $30 a month, which is just a BIT much for something you're expecting to make anywhere from $0-$5 from.

 

[Spacerunner357]

Is there an AI voice that does it in Jive? If so, I'm willing.

Hm Jive? What's that? An I would voice it as well like maybe I can do it like Ciry so I can use it for my books

If I understand their system correctly, I'm 2000 words short of finding out what PocketFM's "reader" (AI) sounds like on one of mine.

Cool i really want to hear what it sounds like like a Military Radio or something
Of a man or And woman

I found one a while ago that was at least mid-quality. But with proper tweaking, the only thing missing was the audible breaths that you hear but don't really pay attention to from a human speaker. It could take direction for more emotional pinpointing, you could use different voices, so you'd have a narrator and a different voice for each character, and you could adjust speed and pitch for each section as well. The issue was that it had a 500 word trial and cost like $30 a month, which is just a BIT much for something you're expecting to make anywhere from $0-$5 from.

Thirty dollars for only 500 words!!!?

 

[lambenttyto]

Audiobooks are expensive to make, and it's probably the only way a lot of self-published authors will ever be able to get into that market.

Yes, and they're a tone of work to produce.

You making any money with your writing?

I'll narrate it for you.

Maybe...

www.youtube.com/@LawrenceCaldwellAuthor

 

[MarekSusicky]

I found one a while ago that was at least mid-quality. But with proper tweaking, the only thing missing was the audible breaths that you hear but don't really pay attention to from a human speaker. It could take direction for more emotional pinpointing, you could use different voices, so you'd have a narrator and a different voice for each character, and you could adjust speed and pitch for each section as well. The issue was that it had a 500 word trial and cost like $30 a month, which is just a BIT much for something you're expecting to make anywhere from $0-$5 from.

As audio books are expensive I was lookin' at the best in AI market - elevenlabs and their prices are steep, but the quality is way ahead the market and ya could do (in theory) tweaking with each voice. But again, the price...

 

[CharlesEBrown]

Hm Jive? What's that? An I would voice it as well like maybe I can do it like Ciry so I can use it for my books

You'd have to check out that classic old documentary comedy, Airplane! for that one...

Thirty dollars for only 500 words!!!?

Unless I misread, the free trial is only 500 words, and after that it is $30/month average.

 

[Clo]

As audio books are expensive I was lookin' at the best in AI market - elevenlabs and their prices are steep, but the quality is way ahead the market and ya could do (in theory) tweaking with each voice. But again, the price...

I use them to help me notice typos my eyes can't read anymore. Elevenlabs isn't terrible, but it's far from great, too.

Still, for an indie person who wants to listen to my own book on walks as a way to find where to improve my writing? I find it a valuable tool.