AI cloned voices in official scenarios

[Zaubermuffin]

Someone just told me about this and I think it's a bit insane. If you go through the official SimRail scenarios and listen to some of the files, you'll find very familiar voices. Including (en/):

• Donald Trump (Ł1_08_carg_ŁP_G2)
• Joe Biden (Ł1_02_tlk_WŁ_Sk2)
• Barack Obama (Ł1_08_carg_ŁP_Ko2)
• Jeremy Clarkson (Krk_04_tlk_KatoKrk_K5)
• Samuel L Jackson (Krk_04_tlk_KatoKrk_Kz5)
• and probably many more that I don't recognize

Some of them are used multiple times; some of them (e.g. Obama, Biden and I think Clarkson) appear also in some of the Polish versions (e.g. pl/Ł1_03_r_ŁS_Sk2.mp3).

I don't mind AI in video games, if it is done ethically (i.e. from proper sources with authorization) and declared as such. But this is neither - I'm pretty sure there's no permission to use those voices, and SimRail isn't listed as using AI on the Steam page, which I think is also a violation of the Steam ToS.

Could we get some kind of official statement on this?

[error723]

Could you share the methodology you used to confirm that these are the “voices” of the individuals you identified? I’d like to know how you verified that these are indeed their voices and not those of other people.

This is a rather serious accusation, and I’d like to be certain it isn’t something made up on a “I know but won’t say how” basis.

Edited December 1 by error723

[Zaubermuffin]

3 hours ago, error723 said:

Could you share the methodology you used to confirm that these are the “voices” of the individuals you identified? I’d like to know how you verified that these are indeed their voices and not those of other people.

This is a rather serious accusation, and I’d like to be certain it isn’t something made up on a “I know but won’t say how” basis.

You can open the files yourself and check. They're MP3 files lying around in the game's directory (e.g. SimRail\SimRail_Data\StreamingAssets\Sceneries\5_Lodz1\Scenarios\08_Cargo-LodzPru\Sounds\en\Ł1_08_carg_ŁP_G2.mp3). The files in particular sound, in my opinion, so astoundingly similar to the real persons - including pronounciation and intonation - that I don't think there's any kind of detailed heuristic necessary.

Edit: To add to that, if this were a one-off (i.e. a single well-known voice) in a sea of otherwise unremarkable ones, I would completely agree with you that further scrutiny is warranted. But if you have a clear pattern ("US presidents"/"famous people") across a dozen or more files, with half a dozen or more different, rather easily identifiable people, it's less about a one-off incident of a rogue developer or something and starts to feel more like this is the standard and done by design.

Edited December 2 by Zaubermuffin

[error723]

5 godzin temu, Zaubermuffin napisał(a):

You can open the files yourself and check. They're MP3 files lying around in the game's directory (e.g. SimRail\SimRail_Data\StreamingAssets\Sceneries\5_Lodz1\Scenarios\08_Cargo-LodzPru\Sounds\en\Ł1_08_carg_ŁP_G2.mp3). The files in particular sound, in my opinion, so astoundingly similar to the real persons - including pronounciation and intonation - that I don't think there's any kind of detailed heuristic necessary.

Am I understanding this correctly? You made a serious accusation of copyright theft based on your own belief that it is copyright theft, and the only evidence is that you think it is the voice of the person you identified? So you have no analysis to support this claim?

[Zaubermuffin]

1 hour ago, error723 said:

Am I understanding this correctly? You made a serious accusation of copyright theft based on your own belief that it is copyright theft, and the only evidence is that you think it is the voice of the person you identified? So you have no analysis to support this claim?

The accusation is impersonation/voice cloning, not copyright theft.

Open up any of the files and listen to them. There's three possible options here:

1. They actually got the last US presidents, some politicians, and actors, and paid them to record some voice lines for SimRail in both English and Polish
2. They actually hired some very talented impersonators to do that
3. They used voice cloning software/TTS to generate those files.

There's one option that's very definitely not happening, one option that is quite odd and highly unlikely, and one option that, given the current craze, seems like a real possibility.

What exactly do you expect me to do? Send the file to realdonald@whitehouse.gov and ask if it's him? Do you really want to shine a spotlight to this issue to those people that would have very plausible reason to actually threaten legal action?

[error723]

12 minut temu, Zaubermuffin napisał(a):

What exactly do you expect me to do? Send the file to realdonald@whitehouse.gov and ask if it's him? Do you really want to shine a spotlight to this issue to those people that would have very plausible reason to actually threaten legal action?

Mate, you’re talking about your personal impressions, not an actual case of voice cloning. You don’t have any audio analysis that would indicate your feeling is correct and that, in this specific case, it is truly Trump’s voice rather than an actor simply mimicking it. I don’t know the scope of satire in your country, but I do know that in Poland there are actors and satirists who can imitate the voices of public figures. And here’s the core point: to confirm the existence of voice cloning, you need to show, based on audio analysis, that the recorded voice is indeed the voice of the person you’re pointing to, rather than relying only on your personal impression that it might be that person’s voice.

From my personal experience, it’s easy to be mistaken, even with someone as distinctive in their speech as Samuel L. Jackson — at work I interact with a person who literally sounds just like that iconic actor. And honestly, when I first had a call with him, I was convinced he was using a voice mixer to achieve that effect, and it was only during a general meeting that it turned out this colleague simply sounds that way.

I know that voice cloning is not only highly problematic from a legal standpoint, but also extremely controversial ethically. However, one thing is certain: you have to demonstrate that voice cloning actually took place in order to accuse anyone of it. By the way, are you sure your impression reflects reality? Are you sure those voices are intentionally meant to imitate the people you’re referring to?

[Zaubermuffin]

I'm not going to argue with someone who's obviously doing it in bad faith.

The files reek of AI. It has all the hallmarks of it: Garbled audio, odd noise, this eerie non-sense pronunciation of words. The timing, the pacing, it's all off. Yes, you could get that with an actor too, but once again let me ask you what's more likely: That they hired a dozen impersonating actors, or just use ElevenLabs or similar. The burden of proof should not be on my side, but on theirs. If they have been actors, then I'm sure they can provide some proof about it, and we can go our merry way. It's easier for them to prove that it isn't AI, than it is for me to prove that it is AI.

If you wish to do an in-depth analysis of the audio, then by all means, please go ahead and do it. The audio is really, honestly clear, however. It's not garbled by radio effects or similar, it's as plain as you can get it, and it's not a case of "sounds an awful lot like X" or "reminds me of X", which I would all accept as "weird, but okay". This definitely isn't.

Have you listened to some of the files, at the very least? Or are you demanding proof and analysis for something that you haven't even bothered to look at yourself in the first place? I listened to the files, I had other people tell me about them, so it's not exactly just me hearing things.

[error723]

20 minut temu, Zaubermuffin napisał(a):

The files reek of AI. It has all the hallmarks of it: Garbled audio, odd noise, this eerie non-sense pronunciation of words. The timing, the pacing, it's all off. Yes, you could get that with an actor too, but once again let me ask you what's more likely: That they hired a dozen impersonating actors, or just use ElevenLabs or similar. The burden of proof should not be on my side, but on theirs. If they have been actors, then I'm sure they can provide some proof about it, and we can go our merry way. It's easier for them to prove that it isn't AI, than it is for me to prove that it is AI.

Yes, I checked the files in the directory you mentioned above, and in that file the narrator’s voice does resemble Trump, but what matters is that I don’t know for sure whether it is actually Trump. Similarly, in another file I could say that the next narrator sounds somewhat like Obama, but is that voice cloning? I don’t know.

21 minut temu, Zaubermuffin napisał(a):

If you wish to do an in-depth analysis of the audio, then by all means, please go ahead and do it. The audio is really, honestly clear, however. It's not garbled by radio effects or similar, it's as plain as you can get it, and it's not a case of "sounds an awful lot like X" or "reminds me of X", which I would all accept as "weird, but okay". This definitely isn't.

Why should I be the one conducting the analysis? Am I the one making the accusation, or are you? In your country, is an accusation equivalent to guilt? Is it acceptable to accuse someone without evidence?

You claim I’m acting in bad faith, yet all I’m asking is that you show on what basis you are making the allegation of voice cloning, which is equivalent to accusing someone of stealing a person’s likeness rights. Where exactly is my bad faith, if I’m explicitly asking you to show how you arrived at the accusation of voice cloning? If you think that my approach, where I clearly state that a personal impression is not evidence, indicates bad faith, then indeed, perhaps it’s better for me to withdraw from this discussion.

So I leave you with this question: is an impression evidence? Is being inquisitive an act of bad faith?

Anyway, thanks for discussion.

[mateusz424]

3 godziny temu, Zaubermuffin napisał(a):

I'm not going to argue with someone who's obviously doing it in bad faith.

The files reek of AI. It has all the hallmarks of it: Garbled audio, odd noise, this eerie non-sense pronunciation of words. The timing, the pacing, it's all off. Yes, you could get that with an actor too, but once again let me ask you what's more likely: That they hired a dozen impersonating actors, or just use ElevenLabs or similar. The burden of proof should not be on my side, but on theirs. If they have been actors, then I'm sure they can provide some proof about it, and we can go our merry way. It's easier for them to prove that it isn't AI, than it is for me to prove that it is AI.

If you wish to do an in-depth analysis of the audio, then by all means, please go ahead and do it. The audio is really, honestly clear, however. It's not garbled by radio effects or similar, it's as plain as you can get it, and it's not a case of "sounds an awful lot like X" or "reminds me of X", which I would all accept as "weird, but okay". This definitely isn't.

Have you listened to some of the files, at the very least? Or are you demanding proof and analysis for something that you haven't even bothered to look at yourself in the first place? I listened to the files, I had other people tell me about them, so it's not exactly just me hearing things.

I don't think any scenario writers record their own voices or those of their friends. 
The most commonly used tools are similiar to this:

https://speechma.com/

[Winger]

Dude, are you trolling, or are you some kind of conspiracy believer? Do you, by any chance, also believe in chemtrails, flat earth or global illuminati domination? Seriously, I can't believe that anyone would throw this kind of accusation here.

[Zaubermuffin]

1 hour ago, mateusz424 said:

I don't think any scenario writers record their own voices or those of their friends. 
The most commonly used tools are similiar to this:

https://speechma.com/

And that's fine - like I've said, I have nothing against properly sourced, and attributed, synthesized voices.

I just don't believe that the ones used in SimRail, by the official developers, and distributed over Steam, match any of those.

23 minutes ago, Winger said:

Dude, are you trolling, or are you some kind of conspiracy believer? Do you, by any chance, also believe in chemtrails, flat earth or global illuminati domination? Seriously, I can't believe that anyone would throw this kind of accusation here.

Listen to the files, then come back again. It's incredibly frustrating to be labelled as some kind of idiot when all you have to do is just look at the, frankly overwhelming, evidence there is.

This isn't a secret easter egg or requires to fiddle with some extractor or anything. You can literally just go to the folder, open the files, and listen to them.

[Antiguedad]

2 hours ago, Zaubermuffin said:

Listen to the files, then come back again. It's incredibly frustrating to be labelled as some kind of idiot when all you have to do is just look at the, frankly overwhelming, evidence there is.

This isn't a secret easter egg or requires to fiddle with some extractor or anything. You can literally just go to the folder, open the files, and listen to them.

You are no kind of idiot. I've heard the files, and at least the voices of Trump and Biden are easily recognizable for me. At first this post looked funny, but now it's very sad to see how people are treating you just for having the guts to expose this issue.

[error723]

3 godziny temu, Antiguedad napisał(a):

You are no kind of idiot. I've heard the files, and at least the voices of Trump and Biden are easily recognizable for me. At first this post looked funny, but now it's very sad to see how people are treating you just for having the guts to expose this issue.

Excuse me, but am I an idiot for saying that empirical impressions, even those based on excellent hearing, do not confirm that this is the voice of the person you claim it is? A voice may be similar, but unless you carry out tests, sound analysis, diction analysis, etc., you cannot be certain. I wanted to find out whether the colleague had examined this issue, and it turned out that he simply has excellent hearing and is making accusations solely on that basis. And now, if you feel so sorry, then perhaps you can prove to me that the voice in this file is Trump’s voice and that it was deliberately created in the form of voice cloning. Unless, of course, you created an account on the forum just to say you feel sorry, because you found a problem that others thought needed clarification first?

I repeat: in Poland as well as in the United States, you are obliged to provide evidence for an accusation, you cannot accuse anyone of illegal actions based on your feelings! And trying to appeal to emotions when you lack arguments does not show a proper approach to the ‘problem’ you claim exists. Because if it really does exist, then present arguments that are something more than just your opinion.

Thanks, and please try to present arguments without pushing your opinion. Feeling and fact are not the same thing.

[mateusz424]

Edited December 2 by mateusz424

[Antiguedad]

2 hours ago, error723 said:

Excuse me, but am I an idiot for saying that empirical impressions, even those based on excellent hearing, do not confirm that this is the voice of the person you claim it is? A voice may be similar, but unless you carry out tests, sound analysis, diction analysis, etc., you cannot be certain. I wanted to find out whether the colleague had examined this issue, and it turned out that he simply has excellent hearing and is making accusations solely on that basis. And now, if you feel so sorry, then perhaps you can prove to me that the voice in this file is Trump’s voice and that it was deliberately created in the form of voice cloning. Unless, of course, you created an account on the forum just to say you feel sorry, because you found a problem that others thought needed clarification first?

I repeat: in Poland as well as in the United States, you are obliged to provide evidence for an accusation, you cannot accuse anyone of illegal actions based on your feelings! And trying to appeal to emotions when you lack arguments does not show a proper approach to the ‘problem’ you claim exists. Because if it really does exist, then present arguments that are something more than just your opinion.

Thanks, and please try to present arguments without pushing your opinion. Feeling and fact are not the same thing.

Ok, I'll show the IA voice recordings to not one but two forensic experts so they compare the IA voice waveforms from the files with the real Trumps and Bidens ones (its a joke). I don't claim there is a problem, and at first glance I thought it was funny, a kind of easter egg. I created an account in the forum two months ago because I bought the game and I liked it. I'm not acusing anyone of anything. But for me its really sad to see somebody being called a  terraplanist or illuminati when I think he is right, at least in the origin of the voices.

Edited December 2 by Antiguedad