If english language only has 44 sounds. How long until we can fakeaudio people? Does this software already exist sort of like fakeapp that can pool through spoken audio, pull the sounds then use them to create new full audio?
>Example: Instead of in Rogue one where they fakeapped young leia into the movie they could really have her sound like a royal british princess?
Source:
dvusd.org
Other urls found in this thread:
deepmind.com
lyrebird.ai
youtube.com
twitter.com
they probably already can, it just isn't consumer technology yet
> it just isn't consumer technology yet
cant we just write it in python? That would work I think. I think I figured out how it could be done but noone has done it yet.
well go do it and make money. don't go on any planes because they kill you that way
>don't go on any planes because they kill you that way
well the fakeapp guy is alive, whats the diff between that and this? This is just audio, it should be way easier since all someone has to do is write it in python.
Google did something like this last year. deepmind.com
It already exists.
Do you niggers even Google? The stupidity of this thread hurts my brain... This is the same as going to a woodworking forum, asking if it's possible to create an electric powered saw [it will be called an "electric saw."]
hmmm well that means its doable.
All we need is python to rip audio like it rips video and then generate that audio into statements.
It would be awesome imagine hearing Luke Skywalker and Princess leia doing a cover of justin beiber and mariah carey song All I want for Christmas is You.
>Google
get out
that does exist already.
somebody once posted videos and screenshots and a slideshow iirc of this a few weeks ago.
can't recall the name though and it's still super new.
some other Cred Forumsfag is also working on an AI that makes women nude by removing their bikinis using an algorithm
exciting times in IT, in that regard
What language has the most sounds, does any language cover all sounds?
I have an affinity for Japanese girls speaking English by the way.
>get out
great movie
>planes
Rich and powerful people use planes the most.
this is a really stupid question
I may be ignorant on the topic but at least I'm not rude.
be more fun to have audio than video, could use it everywhere then from youtube like to make us sound like other people like Jack Black!
2016.
Adobe Voco.
that would be an accent, yes if we write a python script to rip sounds to generate sentences it should do it with the accent intact.
Easy peasy!
don't reply to me ever again lil bitch
I know there are some sounds used by languages that aren't used in other languages which makes it almost impossible for foreigners, but most of those are dead or African so it depends on if you mean all sounds in all languages or all sounds in all relevant languages.
>you will live long enough to have an AI assistant generate speech with a broken English accent
I have a reason to continue.
You have no power here, there's nothing you can do to stop me. Sorry you got upset but it's not my fault.
dont talk to me or my wifes son ever again
>I have a reason to continue.
id like someone to just pop open notepad bang out a few lines of python and BOOM! we got our audio ripper!
Im shocked noone has done that yet but figured out video which on magnitudes is much more difficult.
I'm sure most of the people doing it will still fuck it up, like the idiots plastering celeb faces onto porn actresses that aren't remotely their measurements.
I mean Vocaloid is kind of this, no? They just do't publish how they come up with the product.
The literal state of Cred Forums.
Not to mention the autist who just learned how to run python who made this thread.
i thought they just hire a VA for that stuff?
>Not to mention the autist who just learned how to run python who made this thread.
Clearly he is a master level Python in coding since he ran a GUI of a script.
Never forget that the worlds most elite hackers came from Cred Forums hitting the LOIC button.
Watch out guys we got a super hacker here!
Peaches, I love peaches
Do y'all'd've think that every language has the same consonants?
It's based off a person but the speech is obviously synthesized.
I am waiting for them to reproduce Frank Mullers voice, who narrated the first 4 audiobook of the dark tower, but died in a motorcycle accident and couldn't finish. There is plenty of data from the audiobook he read to make a decent voice I think, but one still must account for ingestion and emphasis. That is what would make it difficult and I would think that would have to be manually tinker with.
Inflection *
"I'm really into vore" -Frank Muller
how is it "obvious"
what if their just adding phase differential or reverb on purpose to a real voice?
is that true?
No. user is saying what could be spoken by Frank muller once what I want is made.
That is speech synthesis, it ceases to be a voice recording when it's modified like that, the original sample is but the product isn't. I'm not just being pedantic either, the significance here is that the samples are produced not replayed, with the amount of variations, if they were pre-recorded samples the install size would be immense so they have to be produced at runtime.
The other obvious tell is that sometimes they use multiple people to produce a single vocaloid.
I think the one you're replying to is just rolling with the joke kek
It's already made.
It was 2 years ago.
It was a play on the "ingestion" typo.
>The other obvious tell is that sometimes they use multiple people to produce a single vocaloid.
interesting, these are things I do not know.
FUCKING my autocorrect still needs some learning. They need a deepautocorrect before doing deepvoice desu
I can't find which one it is because there's a lot of them and sometimes they do contracts for other people which are vocaloidS but not part of VOCALOID the product. The opposite is common in their product (1 voice provider for 2 different vocaloids) the most popular example is Len and Rin.
Still doesn't sound natural though. I think if it followed the same machine learning technique by analyzing the dataset from his old reads it would better.
You don't always want to say the same word with the same inflection every time. It depends on the sentence.
>Still doesn't sound natural though. I think if it followed the same machine learning technique by analyzing the dataset from his old reads it would better.
why cant this be corrected using the same methods we use for animating faces in cartoon and video games? Just use a line simulating attitude that affects expression except in this case inflection.
It would be that easy really when you think about it but noone has figured it out even though Python is available online free.
I agree. I I have a mockup in my mind for how it might work but I wouldn't know how to actual code it since I am not an audio engineer
I been thinking about it like a track layout in adobe with the bar for the attitude adjustment?
what are you thinking?
Basically that. Different filters for adjusting the voice in various ways to tweak it just right.
true......
Been trying to figure this out logically. Maybe like how to determine bold type have 7 word phrases read indicating different inflections until the various emotions can be heard?