xVASynth v2

[Zerg Rush]

xVASynth is an AI tool for generating high-quality voice acting lines using voices from video games. With hundreds of voices supported, across dozens of games, the app provides pitch, duration, and energy control at per-letter granularity.

 

The use of neural speech synthesis leads to natural sounding voices, something which is very difficult to do with more traditional methods involving concatenations of existing data. It also means new vocabulary can be generated, outside of what the voice actors have already read out.

OpenSource and free

https://github.com/DanRuta/xVA-Synth

https://store.steampowered.com/app/1765720/xVASynth_v2/

Widows 10

GPU 2 Gb, (if use GPU mode)

HD 10Gb

 

[Nort]

Too bad it's for Windows 10.

[freyk]

Wel, it should work also for windows 11 or older.  Just try it. It is free. (Its developed using python. So  it can be ported to older or other os'ses)

Wow, thats a nice app. After looking at some demo videos.

But what are audio rights, when using it for TDM?

Demo video:

 

Edited June 11, 2022 by freyk

[Nort]

I already looked into it, and they got it to work on a Windows 7 virtual machine, at which point I just give up. It looks great, but personally I'm not that passionate that I'm going to install virtual OS:es.

[datiswous]

Cool, so I can take voices from TDM to create new voice lines for in TDM. 

[Nort]

5 hours ago, datiswous said:

Cool, so I can take voices from TDM to create new voice lines for in TDM. 

Yes, but as you can see in the demos, even big videogame characters like C&C:s Kane, with lots of voice data, produce fairly choppy lines. If the TDM lines fail, try basing them off of Troy Baker instead, or (for the protagonist) Stephen L. Russell and (for Benny the guard) Daniel Thron.

[datiswous]

I think the curent tdm voice actors might not like to hear their voice say things they didn't say

[Nort]

I'm sure that Mohammed wouldn't like his work to be defiled or burned either, but that typically falls under free artistic expression. Famous people have been exploited in this way since way before deepfakes came along - pretty much since idol posters were a thing.

Sorry.

[Zerg Rush]

I have put this app, because I think that not everyone here has easy access to a voice actor to create dialogue for the missions they create, this is a viable alternative. There is no infringement of possible copyright violations, since, unlike Deep Fake, the original voices are modified, changing tones and adapting it to the text. The AI is only based on one or more of the chosen voices.
This is more similar in voices to what ArtBreeder does with images, not DeepFake.

[freyk]

@admin Maybe movng this topic to our sound subforum would be nice.

[NeonsStyle]

I spoke to Dan about this, and he did say, if someone could provide him with the text of Thiefs vocalisations, he'd be happy to include Thief in the VaSynth.  

[Nort]

40 minutes ago, NeonsStyle said:

I spoke to Dan about this, and he did say, if someone could provide him with the text of Thiefs vocalisations, he'd be happy to include Thief in the VaSynth.  

...and with "vocalisation" I'm assuming that you mean just all the samples from the game, and not some special studio format of them.

[freyk]

Uh, Thief (the player) or Garreth?

How about making "templates" for All the existing voices inside TDM.

[Nort]

About making templates: I've gone over some recent AI voice synthesizers, and one of them requires the voice actor to read a specific long text provided by the program. That sounds like a very thorough way to make a template, if they're willing to do that, which would probably produce good results. I can't remember what synthesizer required that, though. I'm just describing what "making a template" could entail.

Edited June 14, 2022 by Nort

[NeonsStyle]

On 6/14/2022 at 3:51 PM, Nort said:

...and with "vocalisation" I'm assuming that you mean just all the samples from the game, and not some special studio format of them.

He means the transscript of what the game characters say along with the voice files to train the AI. He doesn't use a games actual files other than to listen to to get the AI to sound right. 

[Oktokolo]

On 6/11/2022 at 2:06 PM, freyk said:

Demo video:

The voices are still in the uncanny valley. But they definitely aren't sounding as monotonic and glitchy as in the ole Microsoft Sam times...

It should be possible to get a sortof good sounding voice when feeding it with some weeks worth of audio books or youtube videos featuring only one speaker and no background noise.

[Frank Cotton]

On 6/26/2022 at 6:19 AM, Oktokolo said:

monotonic and glitchy as in the ole Microsoft Sam

take that back and apologize to sam

[Oktokolo]

1 hour ago, Frank Cotton said:

take that back and apologize to sam

Powered by https://tetyys.com/SAPI4/:

Edited June 27, 2022 by Oktokolo

[Hooded Lantern]

IVONA Amy my love. too bad she is with amazon now and better than ever.

Edited June 28, 2022 by Hooded Lantern