/siberia/ - Off-topic

It's le bad because of some petty-bourgeois romanticism uhh I DONT WANT TO BE PROLETARIANIZED NOOOO

>Tell me why this is wrong.
Its not.

AI can't, it's shit at it. However, machine learning has been used for this with great success before AI took all the funding away from actually useful machine learning projects.

>>832679
>AI can't, it's shit at it.
It certainly can but the quality isnt perfect yet. Have to wait a year or two to get perfect human like voice. You can listen samples here:
https://elevenlabs.io/en/use-cases/audiobooks

>>832679
>AI can't, it's shit at it.
i mean i used to listen to textbooks with shitty ass robot TTS. you can't tell me elevenlabs or whatever it's called that every fucker uses to make youtube shorts these days isn't a step up from that.

it can never get all the way there, there will always be something off about it that the listener can't quite put the finger on but will become increasingly conscious of the longer they listen to it. even with the best computer scientists and linguistics and audio engineering experts using all their powers and all their skills they just can't replicate it, because they don't actually know what it is they are trying to replicate. it's one thing to make a convincing emulation of mechanical sound sources like analog synthesizers or acoustic pianos where you know the exact mechanics of how the sound is generated down to the most minute variables, but the human voice is infinitely more complicated than those things - even things like acoustic guitars and wind instruments are beyond our ability to emulate convincingly with computers due to the high degree of biological nuance influencing the sound, nevermind trying to emulate a sound source that is 100% biological. and that's just talking about the pure physics of it - there is also the emotional component in play that neurology has yet to even begin to unravel, we hear human voices or see human faces and the brain unconsciously and effortlessly generates its own extremely detailed subjective interpretations dually influenced by both outside and inside information by some means that is not understood, and then it forms emotional connections to the humans they belong to by some means that is not understood, and we cannot emulate this emotional connection and create virtual human personalities without understanding how the real thing actually works. all we can do is throw together this frankenstein's amalgam of lots of different human faces or human voices blended and mutated together to produce a simulacrum which may look/sound superficially convincing in a particular narrow perspective or context but they never hold up under any significant scrutiny. the brain is so sensitive and so adept at identifying human authenticity that even genuine human beings often fail to really pass the test - i've heard many audiobooks that i just can't listen to because the narrator is so insincere or robotic in their reading that i get annoyed or i get bored and start tuning it out, even human beings are often not good at passing as human.

maybe an ai will be able to replicate someone like scott brick, but i seriously doubt it could convincingly replicate something like this

Just read it out yourself, it's not that hard bro.

>>832707
You cant do other shit while reading. Listening is so much easier although you need to rewind constantly if you didnt focus on some part.

>>832707

maybe OP is vision impaired did you ever consider that no of course not ableist nazi chud

>>832664
>Tell me why this is wrong.
It's not wrong. The problem with AI is it can be used to falsify anything, but there's no way to stop it so we might as well use it. I would rather use a synthesized voice to make an audiobook than use my own voice thereby making my voice available to synthesize.

if it's not worth lending a voice to, it's not worth listening to.