Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Apple Books digital narration

authors.apple.com

Post reply on HN
Print view View original

71–80 of 244 posts

Re: Apple Books digital narration

#71

Post by lvl102 »

kzrdude wrote: post #34
lvl102 wrote: post #5

I am surprised it’s taking so long for big tech companies to roll this out. I suppose there’s too much money in audiobooks? Perhaps licensing issues? If anyone can get a “celebrity reads aloud” feature that would be Apple and that could be big.

The quality needs to be very high for it not to be jarring. I stay with podcasts if the host has a "good radio voice". ("Not even all human voices are good enough for me.") It's just a very intimate medium to have someone's voice right in your head. If the voices have annoying quirks, those audiobooks will not be loved. Yes, it needs to be tens of hours of perfectly good narration.

I think you’re underestimating state of the art in this area. You can do amazing things with just a few minutes of readings.

Re: Apple Books digital narration

#72

Post by leokennis »

cprecioso wrote: post #13
csande17 wrote: post #8

The "Helena" sample contains a pretty good test case of this system's ability to guess where emphasis and pauses go: > I know Bill carried within him deep currents of spiritual yearning that he found easiest to express through the beauty he saw in all places wild. (I checked in an online sample of the ebook: there is no punctuation in this sentence.) Unfortunately the AI completely faceplants, placing an enormous pau…

I guess this is more of a case of garbage in - garbage out. The original sentence itself is not well-structured. But people who write like that and don't edit it; won't care of how the AI reads it. I don't feel like this is a product for carefully producing audiobooks, but to create them by the pound, so to speak. I'd say it's a move for the "make your own business through audiobooks" people [1] -- very strange for A…

To me this feels like the automated video slideshows that Apple Photos (and undoubtedly Google Photos) makes for you. Perfectly fine, but indeed even on a casual watch you notice mistakes/imperfections you'd never make if you were producing such slideshow manually.

But that's the thing...is "perfection" worth 3 hours of video editing for something you casually consume?

I think almost any audiobook listener will vastly prefer a serviceable but imperfect audiobook when compared with no audiobook at all.

Re: Apple Books digital narration

#73

Post by bottled_poe »

dbspin wrote: post #52
ClassyJacket wrote: post #21

Earlier quoted context omitted.

I can understand that sentence fine, I don't see anything wrong with it.

Writer here, it's a terrible sentence. "I know Bill carried within him deep currents of spiritual yearning that he found easiest to express through the beauty he saw in all places wild." "Deep currents of spiritual yearning" is both cliched and unspecific / unclear. "That" is superfluous. It's unclear how Bill is expressing the beauty he sees, and the sentence structure implies he's somehow responsible for the natura…

> Writing is to a large extent the process of expressing a thought or feeling

Neither of which we believe a machine to be capable of conceiving.

Re: Apple Books digital narration

#74

Post by abraxas »

BuckyBeaver wrote: post #16

"Digitally narrated titles are a valuable complement to professionally narrated audiobooks" Yeah, right. What a lame attempt to deflect the (fully warranted) criticism that this will put audiobook narrators out of work.

Some narrators never should have had their jobs to begin with. I'm viscerally angry at the narrator of "Permutation City" on Audible. Such a great book that could not have a more bored, disinterested narrator who clearly doesn't understand the text he's narrating.

An AI TTS engine at this level would do a far better job of it than that particular dude.

Re: Apple Books digital narration

#75

Post by JustSomeNobody »

qwerty456127 wrote: post #68
andy_ppp wrote: post #63

"Once your request is submitted, it takes one to two months to process the book and conduct quality checks." My guess is that these generated voices are far from perfect and someone has to go in and crank the algorithm to get a fair number of passages to not sound strange. Even in the example Helena there is a word at the end of a sentence that sounds like it should be in the middle and has a bit of weirdness to it.…

Why is that we still can't have a perfect or near-perfect text-to-speech given all the astonishing advances in ML taking place? Is TTS an area nobody is really interested in or is it harder than generating beautiful pictures and sophisticated writings? This thing by Apple already sounds way better than the best I heard previously (NextUp Ivona) but it is not an instant-result offline tool yet and that's sad.

My uneducated opinion on the matter is that we are more tolerant of subtle errors in pictures and writings than we are in sounds. Subtle variations of tone can change the meaning of a conversation that words on paper just can't convey.

Re: Apple Books digital narration

#76

Post by justincormack »

qwerty456127 wrote: post #68
andy_ppp wrote: post #63

"Once your request is submitted, it takes one to two months to process the book and conduct quality checks." My guess is that these generated voices are far from perfect and someone has to go in and crank the algorithm to get a fair number of passages to not sound strange. Even in the example Helena there is a word at the end of a sentence that sounds like it should be in the middle and has a bit of weirdness to it.…

Why is that we still can't have a perfect or near-perfect text-to-speech given all the astonishing advances in ML taking place? Is TTS an area nobody is really interested in or is it harder than generating beautiful pictures and sophisticated writings? This thing by Apple already sounds way better than the best I heard previously (NextUp Ivona) but it is not an instant-result offline tool yet and that's sad.

The pictures have weird limbs and the writing has errors. A book is long therefore there will be a lot of issues.

Re: Apple Books digital narration

#77

Post by jtbayly »

dbspin wrote: post #52
ClassyJacket wrote: post #21

Earlier quoted context omitted.

I can understand that sentence fine, I don't see anything wrong with it.

Writer here, it's a terrible sentence. "I know Bill carried within him deep currents of spiritual yearning that he found easiest to express through the beauty he saw in all places wild." "Deep currents of spiritual yearning" is both cliched and unspecific / unclear. "That" is superfluous. It's unclear how Bill is expressing the beauty he sees, and the sentence structure implies he's somehow responsible for the natura…

It doesn’t matter if the sentence is terrible writing. It can still be easily read out loud correctly, but not by this computer program.

Re: Apple Books digital narration

#78

Post by lxgr »

qwerty456127 wrote: post #68
andy_ppp wrote: post #63

"Once your request is submitted, it takes one to two months to process the book and conduct quality checks." My guess is that these generated voices are far from perfect and someone has to go in and crank the algorithm to get a fair number of passages to not sound strange. Even in the example Helena there is a word at the end of a sentence that sounds like it should be in the middle and has a bit of weirdness to it.…

Why is that we still can't have a perfect or near-perfect text-to-speech given all the astonishing advances in ML taking place? Is TTS an area nobody is really interested in or is it harder than generating beautiful pictures and sophisticated writings? This thing by Apple already sounds way better than the best I heard previously (NextUp Ivona) but it is not an instant-result offline tool yet and that's sad.

Are visual generative models really that more advanced, or could this simply be an artifact of their usage?

With generative visual art, people usually spend considerable time fine-tuning the results, and we don‘t get to see all the prompts that didn‘t work out (except if the failure is notable in some way).

Try e.g. illustrating a book, but using only your first prompt for each image. I think the quality would be in the same ballpark as having Siri narrate the corresponding audiobook.

Re: Apple Books digital narration

#79

Post by botro »

dbspin wrote: post #52
ClassyJacket wrote: post #21

Earlier quoted context omitted.

I can understand that sentence fine, I don't see anything wrong with it.

Writer here, it's a terrible sentence. "I know Bill carried within him deep currents of spiritual yearning that he found easiest to express through the beauty he saw in all places wild." "Deep currents of spiritual yearning" is both cliched and unspecific / unclear. "That" is superfluous. It's unclear how Bill is expressing the beauty he sees, and the sentence structure implies he's somehow responsible for the natura…

Could you do Hemingway, McCarthy and Joyce next? It's untenable they wrote 'unspecific / unclear / superfluous / awkward / anachronistic' language such that our robots cannot render to speech.

Re: Apple Books digital narration

#80

Post by andy_ppp »

qwerty456127 wrote: post #68
andy_ppp wrote: post #63

"Once your request is submitted, it takes one to two months to process the book and conduct quality checks." My guess is that these generated voices are far from perfect and someone has to go in and crank the algorithm to get a fair number of passages to not sound strange. Even in the example Helena there is a word at the end of a sentence that sounds like it should be in the middle and has a bit of weirdness to it.…

Why is that we still can't have a perfect or near-perfect text-to-speech given all the astonishing advances in ML taking place? Is TTS an area nobody is really interested in or is it harder than generating beautiful pictures and sophisticated writings? This thing by Apple already sounds way better than the best I heard previously (NextUp Ivona) but it is not an instant-result offline tool yet and that's sad.

Because to understand intonation and rhythm you need to perfectly understand context and emotions. I don’t doubt these things will be added soon enough, so I expect perfect reading end of this year and perfect reading in anyone’s voice with a few samples in 2024.
Post reply on HN