AI-Narrated Audiobooks Now Available in Apple Books

[confirmed]

I’m surprised they didn’t start with Apple News Audio. Seems a good place to test as I don’t think this feature is widely used - and would enable spoken audio features on more pieces, if not all articles available in Apple News.

Maybe in iOS 17?

 

[MrMojo1]

There are already some apps available that can convert and read text from epub, pdf or documents using AI voices for awhile. The AI voices are improving and sound less robotic.

 

[Frantisekj]

It's better than simple TTS, but the robot leaks through a lot. It's extremely low-quality compared to any decent human narrator, but I suppose it's better than nothing. I can't imagine anyone paying actual money for this though.

I use TTS a lot to read books even it fails miserably sometimes in my language. Those voices are fine but only first one sounds natural to me, I think Speechelo voices sounds better.

When you look on conditions to produce book this way I know for sure I wont get anything like that to listen in next decade. English only, romance and fiction only, not compliicated text and so on....

 

[hortod1]

Couple of those samples sound a lot like NPR hosts

 

[dysamoria]

I listen to a lot of audiobooks, but sometimes the voice actors just don't do it for me. Having multiple AI voices to pick as alternatives would be an interesting development. But I'm curious how well they compare dramatically.

Curious, though: "mysteries and thrillers, and science fiction and fantasy are not currently supported" Apparently the voices are trained by genre, so they must be based on existing audiobooks? I hope the narrators get compensation for that.

Vocabulary. There’s no way specialized vocabulary and fictional words (invented names for people and places, etc.) will work without a lot more handholding and custom “training”.

Hell, Siri constantly speaks badly just with common language (I’ve been finding it can’t seem to connect clauses with multiple people named, to other clauses with “and” correctly; one of my recent dictated iMessage replies wouldn’t even register as having happened at all, three times in a row, even though I watched it display on screen as I spoke it the second & third times, as Siri just erased it and asked again & again if I wanted to reply).

There is no artificial intelligence. There are cleverly-written algorithms that seem amazing under very limited contexts, but then fail to work in endless other contexts/circumstances because they’re not capable of thinking. The computer industry has brought us Artificial Stupidity, not AI.

 

[dockgaze]

Okay . . . . so we gain digital efficiency and convenience, and I presume eventually lower cost production thus higher profits etc etc.

And what might be lost? PROSODY - which represents the authors intended emphasis . . . . that is, the subtle intonations and syllabic nuances that convey meaning beyond word definitions. It is not just WHAT we say, it’s also HOW we say it.

So then the question arises WHO programs the linguistic prosody presented in the AI speech? Certainly not the author or a professional voice actor. So yes . . . . we do get word meanings conveyed in a more human-like mode, but we also get a biased and truncated presentation.

Just another dumbed down fake hot rod in my book. You know like, is McDonald’s a real hamburger? I guess, if I’m hungry enough. But would you choose it if given a variety of more inspired choices? Maybe the authors themselves should sign off on the final cut . . . . .

 

[Mousse]

I want a James Earl Jones (Darth Vader) voice or a Morgan Freeman voice. Otherwise, I'd rather read a book myself since it'll be 10x faster than listening.

There is no artificial intelligence. There are cleverly-written algorithms that seem amazing under very limited contexts, but then fail to work in endless other contexts/circumstances because they’re not capable of thinking. The computer industry has brought us Artificial Stupidity, not AI.

There are machines capable of learning like a human, but the processing power required is way, way more than what an iPad or iPhone can handle. I agree that a lot of so called AI is very stupid outside a very narrow range.

 

[captainorange]

Apple has now launched Apple Books digital narration, offering a new way for publishers to automatically generate high-quality AI-narrated audio from written text.

The feature, first announced in December via the Apple Books for Authors webpage, allows publishers on the Apple Books platform to opt-in to have their written books converted into a narrated audio form using AI. Samples of the voices developed specifically for the feature are available on the same webpage.Apple is offering different AI voices for different genres and the feature is only available for some genres at this time, but more will be added in the future. Apple says that it can take up to one month for an AI-narrated audiobook to be created and approved, suggesting that there is an element of manual review in the process. Publishers are also free to offer a traditional, human-narrated audiobook alongside the AI-narrated version.

As highlighted by The Guardian, the first AI-narrated audiobooks are now available in Apple Books, highlighted by the tag "Narrated by Apple Books."

Article Link: AI-Narrated Audiobooks Now Available in Apple Books

I'm a storyteller and children's entertainer. These voices are an upgrade to be sure and I am excited about how much better these sound than a couple of years ago, but they are still limited to a competent 'neutral' reading. A neutral read can be good for news and it is also the kind of thing I might prepare for myself when I am learning a new story and don't want to accidentally encode all my decisions about how to inflect while I learn the relevant text of the story. Even useful in an audio book context when I don't want another person interpreting the author's text for me.

Good reading for fiction--or any form of persuasive speaking--involves not just properly formed words, but context. It is an interpretation and a performance. There are choices about how to cue your listener about how to feel about what you say. The AI is not (yet) at the point where it can comprehend any but perhaps the most basic proximity in terms of relationships to the words--never mind a passive aggressive insult, a clever allusion to a famous work, or an ironic turn of phrase that a reader might recognize from a previous chapter or two books ago in the series.

Until there is some form of comprehension and an algorithmic response that communicates to a listener that the speaker actually understands what is being read and is making interpretive choices, that inflection will be absent or provided by a human who can adjust the AI voice based on context (probably more work than hiring a good performer).

As I said above, there are plenty of use cases for AI reading right now, but a good dramatic performance isn't currently one of them.

 

[zfletch74]

These sound pretty good! But I also can't help but feel a little claustrophobic from not hearing the voice take a breath.

 

[mongobongo]

I was very impressed by all of those voices. Would have never been able to tell that they aren't real. Never listen to audiobooks myself though.

 

[Tagbert]

Man does anyone else remember when Amazon tried this when they first launched the Kindle? It was a ********* of epic proportions and they backpedalled immediately. Guess Apple is big enough that they get to call the shots now.

Amazon’s problem was that because this was a purely client-side feature, they didn’t have copyright for the audio performance of the works. The authors and publishers threatened lawsuits and Amazon had to drop the feature.

Apple is offering this as a service to authors to let them add narration to their books. This does not violate copyright regulations. Different business model.

 

[T'hain Esh Kelch]

It's better than simple TTS, but the robot leaks through a lot. It's extremely low-quality compared to any decent human narrator, but I suppose it's better than nothing. I can't imagine anyone paying actual money for this though.

If it is the difference an audiobook and no audiobook, I could See a lot paying for it. It would also be an excellent and cheap way for indie authors to publish their books as audiobooks.

 

[torontomax]

On guardian’s page ppl were complaining about the art. I think it’s a great thing. Yes replacing people with machines is terrible and would make ppl lose jobs. So did the industrial age. We’re still alive there are still jobs around. I genuinely don’t understand why people are afraid of a man made thing. It’s not like we’re in the dark ages and blaming the ai being the devils work.

Lazy people who are afraid of learning new skills are the ones who holds the torches and stakes.

 

[Syk]

Jackson sounds the best to my ears, still misses some of the little nuisances that you get from a human voice reading.
The Martian is still one of favorite audiobook to listen to

 

[entropys]

I’m surprised they didn’t start with Apple News Audio. Seems a good place to test as I don’t think this feature is widely used - and would enable spoken audio features on more pieces, if not all articles available in Apple News.

Maybe in iOS 17?

Maybe that is why it is mostly limited to romance novels at this point.

 

[Miles Teg]

I see this beneficial for books that have been published before Audiobooks existed. The amount of sales for for an individual title probably doesn't warrant hiring a person to narrate it, but if the AI can do it rather cheaply then this would be great.

 

[The Cappy]

I listen to a lot of audiobooks, but sometimes the voice actors just don't do it for me. Having multiple AI voices to pick as alternatives would be an interesting development. But I'm curious how well they compare dramatically.

Curious, though: "mysteries and thrillers, and science fiction and fantasy are not currently supported" Apparently the voices are trained by genre, so they must be based on existing audiobooks? I hope the narrators get compensation for that.

I wonder if the problem is that science fiction and fantasy is more likely to have totally made-up names and words, and the robots don't handle those reliably enough.

And I agree with you about voice actors. There are a lot of readers where I wish I could just have a computer voice instead. Scott Brick is one of them. A lot of people like him, but to me he just trowels on the trembling anger, and I have a hard time enjoying books he reads. He is probably well suited to Stephen R. Donaldson's stories, but he butchers the spirit of Frank Herbert. Oh well.

 

[jeffehobbs]

Just a flat-out awful idea. Anyone who's spent any amount of time with AI voices (Apple, Google, Amazon) will tell you they're only moderately acceptable for short amounts of text. Any longer than that fatigues the ear.

Apple used to be a company that were the most discerning customers of their own products. C'mon.

 

[Miat]

The samples are very impressive, but certainly carefully picked.
Listening to the longer preview on the iBooks Store, I could definitely make out some robotic and monotonous speech. Still, they're very good, and I'm willing to bet AI narrators will be the industry standard in a few years. (As opposed to fully self-driving cars …)

Anyone who's spent any amount of time with AI voices (Apple, Google, Amazon) will tell you they're only moderately acceptable for short amounts of text. Any longer than that fatigues the ear.

The older my eyes get the more I use both TTS/AI generated speech and human narration.

I can't always pick the difference between artificial and human straight away, but I can always pick it after a while.

Artificial speech lacks the endless subtlety and variation that human speech has, and quickly becomes monotonous and drone like. That never happens with human speech.

It also cannot distinguish the different context-dependent pronunciations of the same spelling ('live', for example). That clearly requires a whole new level of sophistication.

Don't get me wrong. Artificial speech is pretty good, and getting better, and I am grateful for it. It works well for shorter, more 'neutral' pieces, like most news stories. But it has quite a way to go yet before it really works for longer more human stuff, like novels and poetry.

Never say never. But I seriously doubt it will ever fully replace human narration, which can bring the full human experience and understanding to a reading.

 

[BornAgainMac]

Will iJustine publish her book using this technology?

 

[dk001]

A good step forward.
Let me know when I can listen to them on a non-Apple OS.

 

[Apple_Robert]

Pretty good for AI (re: samples).

 

[monster620ie]

there goes humans jobs.

 

[svish]

Good to see this from Apple.

 

[MacMikePro]

I think this would be especially cool if the author could choose different voices for different characters, have the narrator voice be distinctly different from when characters are talking etc. If the AI can figure out which character is talking and do it all automatically that would be amazing.