Jump to content

AI to voice the additional dialogue?


Recommended Posts

Posted (edited)
2 hours ago, Thacobell said:

I mean, the fact that many voice actors have already stated that it makes them uncomfortable should be a factor here. The alternative is to pretend that consent never matters.

It literally doesn't matter, if they have no legal basis for it. If a programmer or writer turned around and complained about some hobby modder making free extensions of the content (that they have already been paid for), fully credited and transparent then they'd be laughed out. No different to VAs who have sold their voice to be used for a character or whatever. 
Now if your complaint was about privacy where they'd been recorded in secret and then something trained on that, you might have an argument. 

 

2 hours ago, Thacobell said:

Because PEOPLE GO THERE. The entire crux of this issue is consent. Whether you think what want to do is benign or not, nobody has the right to dictate what to do with someone's likeness, except for the person themself. What you are comfortable doing with the likeness of another person doesn't matter when the actual person is uncomfortable with what you are doing.

So? How does people making weird dog fanfiction have ANYTHING to do with legit hobby mod content? People are going to go 'there' regardless. Again, the knives in your kitchen drawer COULD be used to commit a crime and people do, are they a problem? The PC you're using COULD be used to send malicious stuff. Audacity COULD be used to cobble together something pretty horrible with someone's voice. etc. etc. 

As to "the right to", that's the thing, I believe they absolutely do, if it falls under free/fair use and transformative work. Usage in the situation that this topic is about (extending BG1 char's vocal dialogue for the additional NPC Project content) absolutely would do that. You also keep saying "many voice actors" blahblah. Exactly WHAT circumstance are they arguing, just when money is being made, anti-AI altogether? Also, some, or ALL of them? What percentage are we talking here, hey? Does 5% speak for 95% etc? Especially if they're misinformed (as many anti-AI arguers are lol). 

"Many" carrier pigeon handlers likely complained when we moved to better communication methods, too. 
Even then, such is irrelevant since we're talking about a FREE HOBBY MOD, here. 


You're also complaining at being told you're using an emotional argument, while:

  • Basing your core reaction on emotive terms and literal feelings such as "gross". 
  • Using objectively (and factually incorrect) emotive terms such as "stealing". (Where "copying" would be better term to argue around). 
  • Appealing to authority by stating "many artists.." blabla as above. 
  • Referencing irrelevant scenarios such as job losses, corporations and and paid work, when we're discussing a free GAME MOD here. (Where the VAs have already been paid for the work they have actually done).
  • Referencing hypothetical situations such as what people could write about doing with dogs (WTF? The NPC project contains none of this). 

Those are objectively "emotional" focused arguments, that do not have any basis in fact. I still also don't see how any of your points (and any law cited) does not apply or is in any case where the exact same OUTPUT would be made using traditional tools (be that a human vocal impression, or computer aided tools such as Audacity etc.). You seem to have a problem with AI specifically, and most of the arguments against seem to revolve around emotion and pure spite lol.

Edited by BobT
Link to comment

Usually when I drop into a thread that's gotten to four pages this quickly, it's to politely rap some knuckles and ask a person or three to behave, but so far this has been relatively peaceful despite some decidedly oppositional views, so thanks to everyone involved.

Second, I'm going to move this into one of the general discussion forums since the topic has broadened beyond BG1 NPC.

What follows are my own views. We're still discussing internally how or if G3 will host mods with AI-generated content.

1 hour ago, BobT said:

It literally doesn't matter

...to you.

That voice actors, individually and/or collectively, have asked me not to do it is sufficient. Whether I have a legal right to do so is secondary. Any noble motivations I might have--that it's a tribute or that I'm trying to honor their work--are also secondary. Whether there's a meaningful philosophical difference between an algorithm and a human iteratively learning to create art is (you guessed it) also secondary.

They've asked me not to do it, and that's enough.

This simple bit of kindness is what has kept the IE modding community going for a quarter century. Nothing is stopping anyone from grabbing one of my mods and putting it on Dropbox or Github except for me politely asking them not to. It's one of the few unifying principles we have, collectively, , which is why it's something that I take very seriously.

Link to comment

I'm very sympathetic to this way of putting things.

I'd add (I've said similar things in internal discussions) that I think comparison with what else we're allowed to do in modding is a bit moot. We're not really allowed to do any of it, in the sense that we probably don't have an independent legal right. (You can gesture at US Fair Use copyright rules, but it's unclear at best what that really covers, and since it's a gray area there's absolutely zero chance of the IE modding community taking on and winning a legal battle with the legit copyright owners.) IE modding persists because (a) it's tiny, harmless and mostly unnoticed by the copyright owners,  (b) insofar as they do notice, they benevolently tolerate it because IE modding is good for the game's brand and supports sales a bit, and more recently (c) because Beamdog actively drew on the modding community's expertise in doing the Enhanced Editions and has made statements actively welcoming modding.

So: the relevant argument can't be "we're allowed to do XYZ, and AI voicing is like XZY, so we're allowed to do AI voicing". It would have to be "none of the copyright owners or other interested parties mind us doing XYZ, and AI voicing is like XYZ, so nobody will mind us doing AI voicing". That doesn't follow logically, and the conclusion clearly isn't true empirically, as Cam says.

Link to comment

why would anyone care what a voice actor wants or doesn't want? their 'wishes' are exactly that - their wishes. none of it is legally binding. unless the va is a personal friend of yours and you don't want to damage that relationship.

in the legal world, ownership and possession ARE legally binding. if the actor owns that voice, they need to prove it (it must be widely known, and it must be recognizable). and i think that's where people are confused. voice actors don't own that asset for bg2. literally, wotc would have licensed that and the actors would have lost all rights to it. so an actor can't go back in time and say 'bg2 used my voice, which i gave, and i want it back'. second, you have to prove that the voice is being impersonated. for example in the music industry, it must sound close enough. but 'close enough' might go to jury, and often has (eg vanilla ice vs queen), midler vs ford, wait vs frito lay. keep in mind, ford and frito lay made money and that has provable damages, which was the main contention in those suits.

"Vocalizations themselves aren't copyrightable, so anyone can do an impression of anyone else, and often the whole point of an impression is to use different words than those used by the famous voice. "

https://www.avvo.com/legal-answers/does-a-comedian-violate-any-copyright-laws-by-doin-311098.html

this is why comedians are covered. in fact, many famous comedians all do impressions. none of them face legal action because again, you must prove damages. just because robin williams does an impression of a celebrity doesn't damage the celebrity's value because people are aware that it's robin williams doing an impression.

AI at least in my country has been declared public use. so anything generated by AI is completely free for all to use. as far as how it's generated, it's fairly complicated. it's a set of linear transforms on what i think is a fourier transform for a lot of stuff like images and audio. when and where this occurs is a bit ambiguous - in fact it could be that only the authors ever knew and now even the authors may not know. which means that the AI would be liable for doing the copying, and since the AI can never make money, it would never be grounds for legal action. this is likely why the AI creators made everything free to use and distribute. the legal trouble, especially in different countries, is basically insurmountable. copyright necessarily requires the making of money in order to be actionable. even auxillary stuff like comedians doing a copy isn't actionable.

"Under U.S. copyright law, voices cannot be protected.  This is because voices are not “fixed” and copyright protection is only available for “original works of authorship fixed in any tangible medium of expression.”[1]  Trademark law also does not protect voices, although sounds that identify goods or services can be registered as trademarks. One example is the “NBC chimes” registered in the Trademark Office (No. 0916522) for television broadcasting services.  Celebrities that have well-known catchphrases can also try to register these as trademarks.  For example, Michael Buffer registered several variations of his famous “Lets’ Get Ready to Rumble!” phrase (e.g., No. 2594218).  Even so, this only protects the phrase, not the voice saying other things."

https://grr.com/publications/hey-thats-my-voice-can-i-sue-them/

and lastly is precedent. the law can't say 'a is true' and then say 'a is also false'. it kind of nulls the entire system. as a result, the first copyright suit against AI will basically set the stage for all other copyright suits. as of yet, it doesn't seem like anyone has pursued this in a court of law in any country. it's very difficult to prove damages. and in fact, the opposite could be argued. if someone posted a mod that said 'new and expanded aerie voice mod', if bg2 made a million sales from it, they couldn't sue that modder. it probably wouldn't make sense to. but again, 'widely known' also needs to be proven, especially looking at medler and wait. i don't think you could prove aerie's voice, minsc's or even sarevok's voices are 'widely known'.

and of course, copyright duration. copyrights are not infinite. i think bg2 has expired its copyrights. there are many legal processes that can be employed to extend copyrights, but i think those doors are soon closing or closed. i have no idea if wotc has extended its copyrights or done any of the legal tactics required to extend copyrights, but the fact that none of the youtube channels posting direct copies leads me to believe that it's no longer actionable. for example, disney has a mouse that recently entered public domain despite their legal tactics.

Link to comment

As the guy literally using AI created voices for an upcoming mod, I'm personally not comfortable from an ethical standpoint of taking an existing voice actor's voice and mimicking it to have them voice something of my creation. I'm not willing to get even close to the court of law on this one either as it is a fresh discussion for people way beyond any of us to hash out from a big picture standpoint. I think Cam said it best: voice actors have expressed their opinion against AI mimicking their voices, and that's good enough for me to honor their request in my work.

Link to comment
6 hours ago, squiros said:

why would anyone care what a voice actor wants or doesn't want?

Read Cam's post.  Your entire argument is legalistic in nature, whereas Cam focused on community norms.  They are NOT the same thing, although far too many people these days conflate them. 

I've been a practicing lawyer for nearly 20 years, and one thing I constantly tell my clients is 'never let your lawyer be your conscience." In other words, just because the law allows for something does not mean you should do it.  Your argument rests entirely on whether or not using AI to mimic voice actors is legally permissible.   That will be decided by the courts.  Cam's argument ignores the legal aspects entirely, and focuses instead on community norms that have kept the IE community going for 20+ years.  

I could get into philosophical discussions about how equating "legal" with "good" is a plague destroying communities all over the world, but that's a topic far beyond this community.  

Link to comment

This was  amajor point of the strike. They wanted to (rightfully) have control over their own likeness. Trying to say that consent is pointless says a LOT about you.

 

Link to comment
On 1/9/2024 at 7:26 PM, BobT said:

It literally doesn't matter, if they have no legal basis for it.

Do you?

No, really. Do you? As far as I know there still no ink over paper law stating you are explicitely, positively allowed to use AI for these ends.

On 1/9/2024 at 7:26 PM, BobT said:

If a programmer or writer turned around and complained about some hobby modder making free extensions of the content (that they have already been paid for), fully credited and transparent then they'd be laughed out.

*looks at fanfic and its larger community agreeing for decades on not daring to ever ask for payment in fear of legal repercussions after some authors made mouth war sounds*

*also at game companies sending Cease & Desists at fangames*

Yeaaah, about that...

On 1/9/2024 at 7:26 PM, BobT said:

So?

So you seem to avoid addressing the consent argument, instead focusing on "might do might bes".

On 1/9/2024 at 7:26 PM, BobT said:

Those are objectively "emotional" focused arguments, that do not have any basis in fact

Oh dear. The "your too emotional" card. This is not the slamdunk you might think it is. And I suggest you read carefully what Cam and Delior are saying.

Edited by Connelly
Link to comment
On 1/9/2024 at 8:39 PM, CamDawg said:

Usually when I drop into a thread that's gotten to four pages this quickly, it's to politely rap some knuckles and ask a person or three to behave, but so far this has been relatively peaceful despite some decidedly oppositional views, so thanks to everyone involved.

Second, I'm going to move this into one of the general discussion forums since the topic has broadened beyond BG1 NPC.

What follows are my own views. We're still discussing internally how or if G3 will host mods with AI-generated content.

...to you.

That voice actors, individually and/or collectively, have asked me not to do it is sufficient. Whether I have a legal right to do so is secondary. Any noble motivations I might have--that it's a tribute or that I'm trying to honor their work--are also secondary. Whether there's a meaningful philosophical difference between an algorithm and a human iteratively learning to create art is (you guessed it) also secondary.

They've asked me not to do it, and that's enough.

This simple bit of kindness is what has kept the IE modding community going for a quarter century. Nothing is stopping anyone from grabbing one of my mods and putting it on Dropbox or Github except for me politely asking them not to. It's one of the few unifying principles we have, collectively, , which is why it's something that I take very seriously.

"That voice actors, individually and/or collectively, have asked me not to do it is sufficient." < Which actors and why btw? Just out of curiosity.

I'd say in general that the legal right to do so is actually primary lol (though I agree that your personal or the team's subjective priorities can be different), people can't just decide others can't lawfully use something because they "don't like" it, else we end up in a world where Apple copyrights the word Apple, etc. The music and movie industries also "don't like" a lot of practices as they'd love to restrict more for more money, same with the "right to repair" and so on. 

That's not saying that I'd not personally listen either of course, however there'd have to be a rational argument behind it, rather than a knee-jerk reaction. (That's the problem I have with a lot of the anti-AI stuff). I'd really love this sort of tool to be used for good as it could be amazing for games, especially remastering or extending old stuff. I think it would be a damned shame to leave it to only be used for malicious purposes because AI, like those knives in your kitchen drawer are a tool, it's what the human does with it that matters. 

However I see that you're listed in the BG1 NPC Project credits (thankyou for your work!) so I'll take this as an official response and where I disagree with some of the premises, I'm happy with the response to the actual question regarding the possibility of using this for the BG1 NPC Project mod by the team, cheers. Happy for your post to also to be treated as the "solved/answered" post if that's possible on here. I've linked it in the OP. 

Edited by BobT
Link to comment
On 1/9/2024 at 9:06 PM, DavidW said:

I'm very sympathetic to this way of putting things.

I'd add (I've said similar things in internal discussions) that I think comparison with what else we're allowed to do in modding is a bit moot. We're not really allowed to do any of it, in the sense that we probably don't have an independent legal right. (You can gesture at US Fair Use copyright rules, but it's unclear at best what that really covers, and since it's a gray area there's absolutely zero chance of the IE modding community taking on and winning a legal battle with the legit copyright owners.) IE modding persists because (a) it's tiny, harmless and mostly unnoticed by the copyright owners,  (b) insofar as they do notice, they benevolently tolerate it because IE modding is good for the game's brand and supports sales a bit, and more recently (c) because Beamdog actively drew on the modding community's expertise in doing the Enhanced Editions and has made statements actively welcoming modding.

So: the relevant argument can't be "we're allowed to do XYZ, and AI voicing is like XZY, so we're allowed to do AI voicing". It would have to be "none of the copyright owners or other interested parties mind us doing XYZ, and AI voicing is like XYZ, so nobody will mind us doing AI voicing". That doesn't follow logically, and the conclusion clearly isn't true empirically, as Cam says.

 

Quote

Do you?

No, really. Do you? As far as I know there still no ink over paper law stating you are explicitely, positively allowed to use AI for these ends.

I'd disagree with this a lot. Law is prohibitive, rather than permissive. Unless it's explicitly prohibited we absolutely DO have the right and it's up to the complainant to PROVE that we do not. It's not just US copyright law, other countries all have free / fair use and transformative work policies. 
Even the VERY heavily regulated music industry allows "covers" of songs and much more lol. Transformative use isn't that a high barrier to clear, sometimes even if the content is the same, just a different USE for the content is enough for it to be "transformative". 

Regardless, it's the knee-jerk and overly emotive reactions I have a major problem with, such as the BG3 Narrator (Amelia Tyler) has been getting alternate narrator mods pulled from Nexus and other sites, yet many of those mods (such as the Male Narrator ones) haven't used her voice in any sort of model, since the model is made for the OUTPUT voice, the man! 

It's going to be better once some court cases get thrashed out and clear a lot of this up I think. There's a prominent one regarding the art side of things where they've mostly gotten thrown out already and the remaining bits are being challenged by the judge to prove that their specific copyrighted work has been substantially used in an output, which I think they've not got a chance of doing. 

Quote

In his dismissal of infringement claims, Orrick wrote that plaintiffs’ theory is “unclear” as to whether there are copies of training images stored in Stable Diffusion that are utilized by DeviantArt and Midjourney. He pointed to the defense’s arguments that it’s impossible for billions of images “to be compressed into an active program,” like Stable Diffusion.

“Plaintiffs will be required to amend to clarify their theory with respect to compressed copies of Training Images and to state facts in support of how Stable Diffusion — a program that is open source, at least in part — operates with respect to the Training Images,” stated the ruling.

 

Quote

If a programmer or writer turned around and complained about some hobby modder making free extensions of the content (that they have already been paid for), fully credited and transparent then they'd be laughed out.

*looks at fanfic and its larger community agreeing for decades on not daring to ever ask for payment in fear of legal repercussions after some authors made mouth war sounds*

*also at game companies sending Cease & Desists at fangames*

Yeaaah, about that...

I specifically mentioned FREE, so why are you bringing up payments? Also anyone can send a cease & desists letter lol, I could send one to you. Doesn't mean I have ANY right to actually make you comply, that would have to be proven in court. Usually though it's a game of bluff and most usually comply as they don't have deep enough pockets to fight it, even if they were to be in the right, large corporations can easily draw a case out to bleed anyone else's pockets dry.

Still though as I mentioned before, it's all about the specific case. Where money or jobs are involved, that's a VERY different situation, and I'm all for creators getting as much protection as possible from exploitation. 

Edited by BobT
Link to comment
22 hours ago, squiros said:

this is why comedians are covered. in fact, many famous comedians all do impressions. none of them face legal action because again, you must prove damages. just because robin williams does an impression of a celebrity doesn't damage the celebrity's value because people are aware that it's robin williams doing an impression.

EXACTLY. And I'd love to see how hobby modders doing a FREE extension to a 26 year old game, fully credited and transparent (especially regarding what's AI and not), could be argued to be damaging in any way lol. Even the jobs / money argument is moot as there was never any additonal work to be paid for, and the VAs had been fully paid for the work they had done already by the modders and every single user, buying the base game. The only additional work done is by the modders. 

There's a lot of emotive work around using "someone's voice", yet that "voice" has already been sold to a corporation for money and now in public and the audio files now owned by the buyers! I'd agree completely to strike off anything that was based on private recordings, though. 
Also I'd love to see how say for example, the VA for Edwin arguing that more of the Edwin voice is using "his" likeness, when it's the character's.. 

Thankyou too for this excellent post, btw. A rational post (though I'm fine if people disagree with your own opinions) primarily based in fact and using case law examples and similarities to existing situations rather than a knee-jerk reaction or emotive arguments, be that for or against. I've linked this in the OP, too. 

Edited by BobT
Link to comment
15 hours ago, Delior said:

Read Cam's post.  Your entire argument is legalistic in nature, whereas Cam focused on community norms.  They are NOT the same thing, although far too many people these days conflate them. 

I've been a practicing lawyer for nearly 20 years, and one thing I constantly tell my clients is 'never let your lawyer be your conscience." In other words, just because the law allows for something does not mean you should do it.  Your argument rests entirely on whether or not using AI to mimic voice actors is legally permissible.   That will be decided by the courts.  Cam's argument ignores the legal aspects entirely, and focuses instead on community norms that have kept the IE community going for 20+ years.  

I could get into philosophical discussions about how equating "legal" with "good" is a plague destroying communities all over the world, but that's a topic far beyond this community.  

Sure! But what's morally "right" is subjective isn't it, and will be different on a case by case basis. 
Would say, someone agree with a politician attempting to silence and prevent someone from performing who does impressions of them? And such impressions are frequently degrading in nature! Now how about when that's completely benign? 
If some sort of reasonable argument is given, I think most would be open to listening, as morpheus562 obviously is. 

Edited by BobT
Link to comment

Also, I'm curious what people think of my friend's have-a-go here: 

Would anyone say that's the Edwin VA's "likeness" and would require his consent? 
She also used her OWN voice to record that, btw. 

I think it's a pretty good start! I can only imagine how good it'd be if done properly and polished.
The SAME result could also be achieved via traditional tools, btw. Albeit slower and less efficient. 

 

Link to comment

I disagree with your points @BobT. There is a vast difference between using a paid service that has its own ai voice actors and algorithms to generate voices (which is what I do) and what you are suggesting which would involve feeding an existing voice actor's work into an ai algorithm to generate new lines with that actor's voice. As I've said before, I view it as a clear distinction between the two, and I will not use AI to copy an existing voice actor.

I personally went through two lawsuits last year (I initiated) and both had favorable results. I also grew up in a home with a big attorney father. In my personal experience, even though you may legally be right, its not worth it being the case study for these things. I would also bet these voice actors and their respective agencies have vastly deeper pockets than you, so even if you are legally right, do you have the funds to go through litigation and prove it? Attorney's fees are not cheap and the big agencies can drag it on and bleed you out. If you somehow come out ahead you still have to suffer the stress, lawyer expenses, and time sink to prove your point (and I'm not sure the law in these types of cases but I doubt you'd be able to recoup your legal expenses).

Link to comment
7 minutes ago, morpheus562 said:

I disagree with your points @BobT. There is a vast difference between using a paid service that has its own ai voice actors and algorithms to generate voices (which is what I do) and what you are suggesting which would involve feeding an existing voice actor's work into an ai algorithm to generate new lines with that actor's voice. As I've said before, I view it as a clear distinction between the two, and I will not use AI to copy an existing voice actor.

I personally went through two lawsuits last year (I initiated) and both had favorable results. I also grew up in a home with a big attorney father. In my personal experience, even though you may legally be right, its not worth it being the case study for these things. I would also bet these voice actors and their respective agencies have vastly deeper pockets than you, so even if you are legally right, do you have the funds to go through litigation and prove it? Attorney's fees are not cheap and the big agencies can drag it on and bleed you out. If you somehow come out ahead you still have to suffer the stress, lawyer expenses, and time sink to prove your point (and I'm not sure the law in these types of cases but I doubt you'd be able to recoup your legal expenses).

Oh I mentioned those points (including your personal subjective decisions, which are fine! Your choice) in my posts above, though there's a lot so I forgive if missed it. 

I also agree there is a difference yes. However how different is the END RESULT to say, opening and tweaking those files in Audacity? Or an impressionist doing the same with one's own voice? 

I also agree (and mentioned) the bit about large corporations etc. being able to bleed dry with dragging out a case, even if they were legally in the wrong. Such weaponisation of law though would hardly help the argument though, would it lol.

Link to comment

Join the conversation

You are posting as a guest. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...