Hey I know a fair few of you aren't that interested with what's going on with our mother-boards but I stumbled onto this thread and definitely thought it's worth talking about
https://boards.4channel.org/mlp/thread/34514258
To sum it all up, anons are in the process of cleaning audio from the series or using newly leaked assets from the show to teach an AI to replicate the voices of the Mane 6, with Twilight being the main focus at the moment. If you saw or listened to the Jordan Peterson deepfakes, it's a similar thing although harder due to the lack of pure clean vocal audio
So far the results are pretty funny
https://vocaroo.com/i/s13Fa6h24wbh
https://clyp.it/u3qsvxas
I'm hoping to give it a try myself although it seems to mess up if the lines are long which is understandable
/mlpol/ - My Little Politics
Archived thread
File (): EE6578DD6E03861883567776AC416E6B-386724.png (377.7 KB, 1280x1280, Science Fiction.png)
File (): DAF4AA67F18CD912327AB27C2CEB3415-9182115.webm (8.8 MB, Resolution:1920x1080 Length:00:00:18, Tw-AI-light.webm) [play once] [loop]
Cool. Good to know that soon we can make Sweetie Belle recite the 14 words.
>>249449
It would be easier (though not cheaper) to just hire a voice impersonator who sounds like one of the VAs. I know some animations have done this.
It would be easier (though not cheaper) to just hire a voice impersonator who sounds like one of the VAs. I know some animations have done this.
File (): 14D83399211DE6FC1368C0A74B79F350-392073.gif (382.9 KB, 470x471, happy pony.gif)
Ok so I've had the chance to give it a go now and I do have some interesting results.
https://vocaroo.com/i/s0gErUETpNaz
https://vocaroo.com/i/s0bm85AOcEF2
The difference between these two is that the first is made from using Hit Learn and the second is straight up Hitler. I do like the tone she has in the second even if its not as clear
https://vocaroo.com/i/s0DgAIntznSF
And I found the best way to get her to say anon is by putting it as An Non
Overall in its current state it can't handle large sentences and while it does end rather abruptly on text given, you can get some believable results so long as you try to stay along lines Twilight does say in the show
https://vocaroo.com/i/s0cAXYtXZTsS
Also this one is so wholesome and by far the best I've done so far
https://vocaroo.com/i/s0gErUETpNaz
https://vocaroo.com/i/s0bm85AOcEF2
The difference between these two is that the first is made from using Hit Learn and the second is straight up Hitler. I do like the tone she has in the second even if its not as clear
https://vocaroo.com/i/s0DgAIntznSF
And I found the best way to get her to say anon is by putting it as An Non
Overall in its current state it can't handle large sentences and while it does end rather abruptly on text given, you can get some believable results so long as you try to stay along lines Twilight does say in the show
https://vocaroo.com/i/s0cAXYtXZTsS
Also this one is so wholesome and by far the best I've done so far
>>249436
Whatever happens, we can NOT, under any circumstances, ever use this technology to make recordings of Twilight Sparkle's voice reading passages from Mein Kampf, /pol/ copypastas, and Donald Trump speeches, and then troll Tara Strong by posting links to the recordings on her Twitter and accusing her of saying those things herself. That would just be unconscionable and wrong.
Whatever happens, we can NOT, under any circumstances, ever use this technology to make recordings of Twilight Sparkle's voice reading passages from Mein Kampf, /pol/ copypastas, and Donald Trump speeches, and then troll Tara Strong by posting links to the recordings on her Twitter and accusing her of saying those things herself. That would just be unconscionable and wrong.
>>249700
Same goes for using this technology to replicate Twilight's voice in fanworks.
It'd just be wrong to deprive the "I sound somewhat like x character" women of their jobs!
Same goes for using this technology to replicate Twilight's voice in fanworks.
It'd just be wrong to deprive the "I sound somewhat like x character" women of their jobs!
As much as I would love to get into this, I promised myself that I wouldn't return to 4cuck after learning how absolutely fucked the moderation is. Those IRC leaks, the whole kicking various things off of the site, and so on, and so on.
For those that do still go there, perhaps you could suggest other Tara Strong voice IPs for more data if it hasn't already been suggested? I think Lollipop Chainsaw cheerleader was similar to Twilight?
For those that do still go there, perhaps you could suggest other Tara Strong voice IPs for more data if it hasn't already been suggested? I think Lollipop Chainsaw cheerleader was similar to Twilight?
>>249736
>other Tara Strong voices
I'm not really knowledgeable on this stuff, but my concern with this is that similar sounding characters could possibly fuck up Twilight's inflections.
>other Tara Strong voices
I'm not really knowledgeable on this stuff, but my concern with this is that similar sounding characters could possibly fuck up Twilight's inflections.
>>249743
True. I know all her characters sound similar but this is to preserve Twilight's voice, not Tara's.
Can you get her to say the Yoshikage Kira copypasta?
True. I know all her characters sound similar but this is to preserve Twilight's voice, not Tara's.
Can you get her to say the Yoshikage Kira copypasta?
>>249689
Hey sorry for the late reply but I'll post the guide here for the most recent usable version.
https://pastebin.com/S5pejVnU
Basically all you need to do here is download all these files and then place them into a folder, from there you need to hold shift and right click on the folder you put all the files in and select "Open Powershell window here" (Or Command Prompt, both still work).
Once you've done that you'll need to paste the part in step 4 into the window (Copying the text and then right clicking into the window will let you paste text that way).
After, just change the "This is a test" and "test.wav" with what you want your text to be and what its called and hit enter, after a few seconds it should have created a file in that folder of your new sound file for you to listen to.
Hope that's helpful and while its not that user friendly its understandable to a script kiddie at least. While I was at it I also made a new one in relation to the season finale
https://voca.ro/7TJDzIh7axI
>>249700
Your mind is telling you no, but your dubs are telling me yes
https://voca.ro/4zzX0zoZVuP
https://voca.ro/94PQxKKw1uz
>>249749
Sadly Twi AI really struggles with large text volumes and so will suddenly turn to incoherent mumbling. But maybe you'd stand a chance if you cut the text into different files and then stitched the audio back together
https://voca.ro/3kJE4SxFhj1
Hey sorry for the late reply but I'll post the guide here for the most recent usable version.
https://pastebin.com/S5pejVnU
Basically all you need to do here is download all these files and then place them into a folder, from there you need to hold shift and right click on the folder you put all the files in and select "Open Powershell window here" (Or Command Prompt, both still work).
Once you've done that you'll need to paste the part in step 4 into the window (Copying the text and then right clicking into the window will let you paste text that way).
After, just change the "This is a test" and "test.wav" with what you want your text to be and what its called and hit enter, after a few seconds it should have created a file in that folder of your new sound file for you to listen to.
Hope that's helpful and while its not that user friendly its understandable to a script kiddie at least. While I was at it I also made a new one in relation to the season finale
https://voca.ro/7TJDzIh7axI
>>249700
Your mind is telling you no, but your dubs are telling me yes
https://voca.ro/4zzX0zoZVuP
https://voca.ro/94PQxKKw1uz
>>249749
Sadly Twi AI really struggles with large text volumes and so will suddenly turn to incoherent mumbling. But maybe you'd stand a chance if you cut the text into different files and then stitched the audio back together
https://voca.ro/3kJE4SxFhj1
>>249796
If you have her read the copypasta in 6-second chunks, does she awkwardly stop at the end of every chunk?
Could a video maker edit the voice clips together for a longer sentence?
If you have her read the copypasta in 6-second chunks, does she awkwardly stop at the end of every chunk?
Could a video maker edit the voice clips together for a longer sentence?
This is amazing!
But why does her voice sound so breathy and on the verge of tears so often?
But why does her voice sound so breathy and on the verge of tears so often?
>>249451
why not just kidnap the original voice cast qnd keep them tied to radiators in your basement forcing them to recite lines if they want to be fed?
why not just kidnap the original voice cast qnd keep them tied to radiators in your basement forcing them to recite lines if they want to be fed?
>>249801
You would just have three overlapping words at the end of the last part and beginning of the next part, (so it doesn't weirdly change the diction), and then use some kind of crossfade cut.
This is literally nigger tier editing 101.
>>249805
I would assume because that's a pattern that recurs in the samples, probably just good breathing support.
You would just have three overlapping words at the end of the last part and beginning of the next part, (so it doesn't weirdly change the diction), and then use some kind of crossfade cut.
This is literally nigger tier editing 101.
>>249805
I would assume because that's a pattern that recurs in the samples, probably just good breathing support.
>>249808
Who needs to do that, and deal with cleanup when you just need 5 seconds of audio.
https://youtu.be/0sR1rU3gLzQ
By gosh voice acters and politions have many many seconds of speech.
The not quite AI revolution is nearing for your voice. It's not perfect, but it's getting better.
Who needs to do that, and deal with cleanup when you just need 5 seconds of audio.
https://youtu.be/0sR1rU3gLzQ
By gosh voice acters and politions have many many seconds of speech.
The not quite AI revolution is nearing for your voice. It's not perfect, but it's getting better.
Imagine, within a few years we could have programs that create audiobooks of erotic fiction, with AI copies of the VAs reading their lines.
Hell, think of the potential for porn games, paired with AI conversation generators. The time is coming soon where we can generate natural dialogue on the fly, and if the program has a voice engine built-in, you could basically talk to your waifu without the need for shitty pre-written dialogue trees and expensive fan VAs.
Hell, think of the potential for porn games, paired with AI conversation generators. The time is coming soon where we can generate natural dialogue on the fly, and if the program has a voice engine built-in, you could basically talk to your waifu without the need for shitty pre-written dialogue trees and expensive fan VAs.
nice
File (): 018228E43CD1A6DC40194F7DE83BE08C-235759.png (230.2 KB, 441x491, 1573691329999.png)
>>249436
Oh, didn't see that thread.
I'm the faggot who did OP's vid, in case you have questions specifically about it.
Oh, didn't see that thread.
I'm the faggot who did OP's vid, in case you have questions specifically about it.
>>251211
Oh are you working on the pony game from the video? Maybe you should make a thread about it? I'm sure some many would be interested, especially the anonfilly crowd from what I've seen
Oh are you working on the pony game from the video? Maybe you should make a thread about it? I'm sure some many would be interested, especially the anonfilly crowd from what I've seen
>>251220
Yep.
Not sure it deserve a thread, I work on it on my spare time so it's not like there is new stuff to discuss often.
I tend to just drop a video in Derpibooru, unless the new thing is related enough to an existing thread.
Yep.
Not sure it deserve a thread, I work on it on my spare time so it's not like there is new stuff to discuss often.
I tend to just drop a video in Derpibooru, unless the new thing is related enough to an existing thread.
File (): 1592EB2D01BA8648AAA8216E814ABA39-2956645.gif (2.8 MB, 285x375, 1511974568011.gif)
>>251223
Love the "lipsync" (or what it is called) animation, and I know this is way outside of topic, but can you add an "lipsync" animation for biting the underlip (not sure how it would look)?
>pic related
>Not sure it deserve a thread, I work on it on my spare time so it's not like there is new stuff to discuss often.
We have threads that are over two years old still going so a low update count is not necessary a hindrance for a good topic.
Love the "lipsync" (or what it is called) animation, and I know this is way outside of topic, but can you add an "lipsync" animation for biting the underlip (not sure how it would look)?
>pic related
>Not sure it deserve a thread, I work on it on my spare time so it's not like there is new stuff to discuss often.
We have threads that are over two years old still going so a low update count is not necessary a hindrance for a good topic.
File (): 3E0CE32292338BCEA2AA1DDE61BD0114-146063.jpg (142.6 KB, 1149x501, Anonfilly try the sexy underbite thing.jpg)
>>251226
... not sure what you mean.
If you just mean "facial expressions", that's what blendshapes (or flexes, depending on software) are.
The SFM models do have a special blendshape for underbite - which look kinda retarded by itself ... maybe an actual animator can mix it with other things to make it look cute.
If you mean, "integrate this kind of expression with the dialogue" then ... not automatically. The PPP guys themselves haven't really tackled emotion and intonation yet.
You can make a classic face animation and blend it with the lipsync automated stuff tho.
... not sure what you mean.
If you just mean "facial expressions", that's what blendshapes (or flexes, depending on software) are.
The SFM models do have a special blendshape for underbite - which look kinda retarded by itself ... maybe an actual animator can mix it with other things to make it look cute.
If you mean, "integrate this kind of expression with the dialogue" then ... not automatically. The PPP guys themselves haven't really tackled emotion and intonation yet.
You can make a classic face animation and blend it with the lipsync automated stuff tho.
33 replies | 6 files | 19 UUIDs | Archived