Human Read vs Computer Generated (TTS) Audiobooks

Everything except LibriVox (yes, this is where knitting gets discussed. Now includes non-LV Volunteers Wanted projects)
annise
LibriVox Admin Team
Posts: 38680
Joined: April 3rd, 2008, 3:55 am
Location: Melbourne,Australia

Post by annise »

I don't think anyone is saying they don't have their uses and they are a big help for those with reading difficulties and visual problems. but we will not host them, that is not what we are.

Anne
SowasVon
Posts: 205
Joined: January 24th, 2022, 5:00 pm

Post by SowasVon »

annise wrote: October 5th, 2022, 5:46 am I don't think anyone is saying they don't have their uses and they are a big help for those with reading difficulties and visual problems. but we will not host them, that is not what we are.

Anne
Yeah, that's how I understood it. I just wanted to tell what my experience with TTS is. It wasn't meant as an argument for its use here - sometimes, I just like talking about stuff, so to speak.
"You're on Librivox? Pffft. You just like to hear yourself talk."
"Yuuuup." :mrgreen:
Alister
Posts: 377
Joined: June 17th, 2022, 9:22 pm
Location: Western Australia

Post by Alister »

I've got a bit of a non-sequitur that I'm curious about regarding computer-generated text rather than voices.

My understanding of copyright is that the work of animals and AI is uncopywritable.

So if a public domain text was translated from one language to another with Google translate, chatGPT or the next big thing is it still Public Domain and is it fair game for LibriVox?

Advances in AI could be big if so.
Rapunzelina
LibriVox Admin Team
Posts: 17801
Joined: November 15th, 2011, 3:47 am

Post by Rapunzelina »

Alister wrote: February 15th, 2023, 11:46 pm is it fair game for LibriVox?
Not really.
https://wiki.librivox.org/index.php?title=Recording_%26_Text_Policies#Texts_that_can_be_recorded_for_LibriVox_include_the_following:
"Traditionally published" is the keyword here.

I don't know about the copyright status though. But I see that in the UK AI works are protected for 50 years.
jmac698
Posts: 55
Joined: December 11th, 2012, 1:46 pm
Location: Canada

Post by jmac698 »

I came here because I was curious as to people's impression of computer voices. I may be one of the few to know about this, but there is a way to add emotion to a computer voice. There is a standard notation called SSML which can change the pauses, pitch, speed and volume for words. Making these changes by hand is of course, a bit harder than just speaking them, but you can duplicate the emotion of a human. I did a test by cloning my voice then adding the same emphasis I used. The goal was to make my voice consistent for different characters. There are also some services which make it easy to just mark a word for emphasis, and there are services which automatically add emotion as well.
ref.
https://en.wikipedia.org/wiki/Speech_Synthesis_Markup_Language
czandra
Posts: 3190
Joined: February 13th, 2021, 1:43 pm
Location: Quebec, Canada

Post by czandra »

SowasVon wrote: October 5th, 2022, 3:47 am I wonder whether a time will come where AI can do as good a job as human readers. Who knows. Not yet, at least.
I don't think AI will weep at the end of a poem like I did last week. Does AI get a lump in its throat that's audible? Or a gleam in its eye that you can hear?

Czandra
I asked my librarian about the noise, and she said, "no one would come here
if they weren't allowed to talk out loud." So I read out loud.

Je lis à haute voix car refléchir fait trop de bruit!
lightcrystal
Posts: 1255
Joined: October 22nd, 2021, 10:55 pm
Location: Melbourne with kangaroos

Post by lightcrystal »

czandra wrote: March 27th, 2023, 1:38 pm
SowasVon wrote: October 5th, 2022, 3:47 am I wonder whether a time will come where AI can do as good a job as human readers. Who knows. Not yet, at least.
I don't think AI will weep at the end of a poem like I did last week. Does AI get a lump in its throat that's audible? Or a gleam in its eye that you can hear?

Czandra
Subtle sarcasm.
Subtle demands of a text. Like one that I have right now where there's no punctuation to suggest it. But the author is saying that he does X then No! A description that he does the opposite.

AI has come a long way from the Microsoft SAM of 10 years ago with the same annoying voices. But I can't see it doing what we can do.
Fan of all 80s pop music except Meatloaf.
GettingTooOld
Posts: 416
Joined: October 19th, 2021, 3:28 am

Post by GettingTooOld »

I'm a fan of both in a way, I'm a total groupie for human read stuff, expecially when it's well done. For trivial things that nobody could possibly be expected to read, where it's not even really worth an audiobook but is just the random thoughts or trivial list of practice tutorials or some such, I wouldn't mind using OFFLINE AI tools to generate them. Which do exist, and are quite good. From the time this thread was started, until it ends due to the EMP of world war three, or you know, an admin locks it, which is the same as WW3 :lol: AI has come a long way. AI is already being used to make it's own kill humans / don't kill humans decisions and carrying them out, and making delightful reading voice clones of random humans.

When this thread started Text to speech (TTS) was like, ergh, much more robotic and easy to spot than it is now. It now sounds quite a lot better. You can get comparsions online of course, I won't give examples. I have been looking at them, but humans are always always better.

Working with humans is much more enjoyable. Just ask the workers being injured and killed in logistic warehouses by pick'n'place bots. It's a read.
BengtW wrote: August 26th, 2021, 10:05 am Since I manage our YouTube channel I get a lot of comments on our recordings.
I'm wondering if there are ads being shown outside of the people google knows are librivox members/workers/readers, cause people mentioned and I read somewhere and so on, about Creative Commons Non-commercial (CC-BY-NC) content IS monetized, as in there are ads and you are the charity and google keeps all the dosh. And you get nothing. But there are ads. So I wonder if they are scalping off the Public domain (PD / CC0) too...

https://www.youtube.com/watch?v=oHRzn2ScWco

It's somewhat common where big companies take volunteer-created work and monetize it, as does amazon with LV. although some get their hands smacked

https://duckduckgo.com/?q=sony+steals+blender+movie&t=keltanang&ia=web

I hate ads so much that I even write my own youtube channel downloader that takes one video address and goes and fetches all the descriptions and thumbnails (full size HD though) and assembles it all into my own local static webpage on my hard drive to be browsed offline and select a list of stuff to download later. without ads. phew. can't stand it otherwise, as I pay for my internet and can't tolerate rich people stealing my bandwidth to rip me off. I get progressively more livid as their attempts to force me to watch escalate.

Working with humans on a good cause is always rewarding and always will be. Long live nice people.!
annise
LibriVox Admin Team
Posts: 38680
Joined: April 3rd, 2008, 3:55 am
Location: Melbourne,Australia

Post by annise »

LibriVox recordings are human-read and public domain in the USA .
Anyone is quite at liberty to set up a site and produce machine-read books and run ads and anything else they want to do and if the time comes when nobody enjoys reading aloud then we won't be producing any more books.

And your new site will be the place to conduct these discussions - I'm not going to lock this at present but it is becoming more off-topic for LV every post so I'll move it there.

Anne
Locked