[Snowcatcher]

Title basically, new guy here, cant express my gratitude to this community, brings back childhood vibes and tears... and some sort of reminiscent feeling of something that once was. Few years back i would glance at my past and think about this game and the fact that this jewel can never be repeated due to the various circumstances such as your youth, the time this was set in, the voice actors... or can it ?

Can we get someone with more time on their hands to do full fledged missions using pre-trained AI voice model feeding on various original voice actors from actual lines in the game ? We can get them to say anythin we want now, and preserve that original tone of voice of Samsa, Petrarch, Bosch, hell why not even the translated Vasudan voice.

Lets go community, we have the gift of AI in our hands, you know what must be done

[mjn.mixael]

You mean aside from the fact that recreating actors digitally without their consent is ethically questionable and often against what they or their living estate would want?

Aside from that little issue, AI tools are never the magic bullet that they are advertised to be. They take serious effort to get the results you want and even then they still draw you a woman with 16 fingers. And yes there is an audio version of that.. it's often weird cadence with a stilted performance. Ask Darius who uses AI generated voices for all of Solaris (not dead-actor recreations, but just generally available AI voices) and how much effort it takes to get each line right.

[deathspeed]

While I appreciate the ideas, I'm not sure AI voicing is quite at the drop-in point yet.  More significantly, I question the ethics and possibly legality of using AI to recreate the voices of real actors.  I doubt HLP/SCP wants that kind of attention if someone (or their estate) objects.   

edit: mjn.mixael already said it better than me lol
edit2: typo

[Snowcatcher]

Yes, I understand both the legal, and moral implications here. However:

Scientists assured us that we have all the data we need to rebuild the Kn...  new community made missions and have realistic briefings by the original voice actors in the FS2 universe.

[0rph3u5]

I won't repeat the legal and primary ethical argument against AI recreations, but there is another point that needs to be made:

The recreation of know talent by AI is raising the entry hurdles for new and developing talents, similarly to the preferential hiring of same famous talents over and over.  This is not only a problem for voice acting talent but also for the producing side, as the variety and diversity of talent can be throttled by this.

[ShadowsOfLight]

Coming out of lurking for a rare post, because I think the landscape around Generative AI is going to get insane over the next few months. I fully recognise that the post sounds over the top and is a total soapbox - and if someone can snap me out of it with a good rebuttal I will be grateful.

Firstly, welcome to the community Snowcatcher! Your idea clearly comes from a good place, that I think we can all appreciate. Freespace had a great impact on all of us, and we want to keep a good thing going. There is huge potential in AI, and already a lot of tangible tools that would have been completely unthinkable even 3 or 4 years ago. I don't in any way want to stifle the enthusiasm and positivity you are showing, which are awesome, but I want to address this topic in a more general sense. The post gets a little deeper than what you brought up, because I think we're all going to be facing increased dilemmas in this space in the coming couple of years.

Good posts already about ethics, legality, and damage to entry level workers

The Knossos is actually a great analogy. It had the power to connect humanity back to roots that it had lost, and make humanity feel whole again. But it also opened the door to a second Shivan Incursion that threatened humanity's very existence.

We have this great AI Voice Acting technology - it still comes with its own workload, but the tools will continue to improve and eventually it will become pretty straightforward. Cool. But why stop there?
Using something like ChatGPT, we could generate the command briefing text and dialogue. In fact, we could go one step further and generate entire campaign ideas, which can be filled in with mission-by-mission scripts. Video AI is hitting its stride this year, so I'm sure we'll get to a point of being able to create command briefings and even full cutscenes this way. There are actually 3D model generators out there in beta testing right now, so very soon we could begin to add an endless variety of new ships. And I've just heard the first convincing examples of music generated by AI (including lyrics and cover art, by the way!), so pretty soon we can add endless soundtrack options to the equation (which is the field I have enjoyed contributing to the community).

Sooner or later, I'm sure LLMs will get to a point where they can churn out a proper FS2 mission file, and tables, given a few examples. Actually, I just checked this - Microsoft Copilot can already generate something that is in the ballpark; though far from perfect, and quite a simple example, I haven't even tried to tighten my prompting or look for a more specialised AI to do this sort of thing.

Given the above, in the near-ish future anyone will be able to independently churn out an epic campaign in a week, as close or as far from the FS vibe as they want. That truly sounds awesome... I mean, I have had a couple of campaign ideas floating in my head for decades now, and I could finally see them happen.

Then we'll basically have endless new FS content... I don't know about any one else, but my free time hasn't been growing lately. So I won't even begin to touch all of the content. Maybe I won't ever touch anyone else's content - after all, I can make exactly the campaigns I want to play, with all of the content I need to keep it fresh. And forget about contributing music to anyone else's campaigns, I would just slow them down or get in the way of their own creativity. No more theorising about lore, or discussion about campaigns, because no one will be playing the same stuff anymore.

Meanwhile, AI is doing the same thing in many areas of intellectual labour. Junior programmers get cut out, marketing teams are eliminated, Hollywood starts churning out fully AI generated films, IT departments are cut by 80%, etc. No one will have time, energy or spare money for hobbies anymore because we're all fighting to be underpaid and overworked by a limited number of jobs. Good thing our entertainment will come cheaper, more personalised and with even less effort than before.

We'll all have to think critically about where we are going to draw lines, not just in this community but in almost everything we are involved in. To be fair, the FS community is small enough that it feels more justified to use these tools than it might in other circles... But we have to be careful we don't get too caught up exploring the other side of the Knossos, otherwise the jump node will stabilise and there will be no escape. (It may already too late.)



With that existential rant out of the way... Some of the existing community content is amazingly well done, so before you get too caught up in the desire for new adventures from old characters, make sure you check out the new adventures from new characters.

[Colonol Dekker]

Coming out of lurking for a rare post, because I think the landscape around Generative AI is going to get insane over the next few months. I fully recognise that the post sounds over the top and is a total soapbox - and if someone can snap me out of it with a good rebuttal I will be grateful.

Firstly, welcome to the community Snowcatcher! Your idea clearly comes from a good place, that I think we can all appreciate. Freespace had a great impact on all of us, and we want to keep a good thing going. There is huge potential in AI, and already a lot of tangible tools that would have been completely unthinkable even 3 or 4 years ago. I don't in any way want to stifle the enthusiasm and positivity you are showing, which are awesome, but I want to address this topic in a more general sense. The post gets a little deeper than what you brought up, because I think we're all going to be facing increased dilemmas in this space in the coming couple of years.

Good posts already about ethics, legality, and damage to entry level workers

The Knossos is actually a great analogy. It had the power to connect humanity back to roots that it had lost, and make humanity feel whole again. But it also opened the door to a second Shivan Incursion that threatened humanity's very existence.

We have this great AI Voice Acting technology - it still comes with its own workload, but the tools will continue to improve and eventually it will become pretty straightforward. Cool. But why stop there?
Using something like ChatGPT, we could generate the command briefing text and dialogue. In fact, we could go one step further and generate entire campaign ideas, which can be filled in with mission-by-mission scripts. Video AI is hitting its stride this year, so I'm sure we'll get to a point of being able to create command briefings and even full cutscenes this way. There are actually 3D model generators out there in beta testing right now, so very soon we could begin to add an endless variety of new ships. And I've just heard the first convincing examples of music generated by AI (including lyrics and cover art, by the way!), so pretty soon we can add endless soundtrack options to the equation (which is the field I have enjoyed contributing to the community).

Sooner or later, I'm sure LLMs will get to a point where they can churn out a proper FS2 mission file, and tables, given a few examples. Actually, I just checked this - Microsoft Copilot can already generate something that is in the ballpark; though far from perfect, and quite a simple example, I haven't even tried to tighten my prompting or look for a more specialised AI to do this sort of thing.

Given the above, in the near-ish future anyone will be able to independently churn out an epic campaign in a week, as close or as far from the FS vibe as they want. That truly sounds awesome... I mean, I have had a couple of campaign ideas floating in my head for decades now, and I could finally see them happen.

Then we'll basically have endless new FS content... I don't know about any one else, but my free time hasn't been growing lately. So I won't even begin to touch all of the content. Maybe I won't ever touch anyone else's content - after all, I can make exactly the campaigns I want to play, with all of the content I need to keep it fresh. And forget about contributing music to anyone else's campaigns, I would just slow them down or get in the way of their own creativity. No more theorising about lore, or discussion about campaigns, because no one will be playing the same stuff anymore.

Meanwhile, AI is doing the same thing in many areas of intellectual labour. Junior programmers get cut out, marketing teams are eliminated, Hollywood starts churning out fully AI generated films, IT departments are cut by 80%, etc. No one will have time, energy or spare money for hobbies anymore because we're all fighting to be underpaid and overworked by a limited number of jobs. Good thing our entertainment will come cheaper, more personalised and with even less effort than before.

We'll all have to think critically about where we are going to draw lines, not just in this community but in almost everything we are involved in. To be fair, the FS community is small enough that it feels more justified to use these tools than it might in other circles... But we have to be careful we don't get too caught up exploring the other side of the Knossos, otherwise the jump node will stabilise and there will be no escape. (It may already too late.)



With that existential rant out of the way... Some of the existing community content is amazingly well done, so before you get too caught up in the desire for new adventures from old characters, make sure you check out the new adventures from new characters.

I enjoyed reading this.

[Kie99]

Morally and legally I don't see it as fundamentally different from what we've already been doing.

For 25 years this community has been using the intellectual property of the writers and developers of FreeSpace.  The guys who came up with the FreeSpace story aren't getting paid when you play a new campaign that's heavily reliant on the story they wrote, the guys who did the modelling aren't getting paid when you have a new campaign with a Deimos in it, or a hi-poly version of their design.  I don't see why voice work is meaningfully different.  It's not like we're going to pay any of the original VAs to do it if we can't do it with AI, nor pay anyone else.  There's no material loss.

The main difficulty would be practical.  The vocals you hear on YouTube are self-selecting for those that are really good.  If you actually try to do it yourself it's much tougher than you might think, the tone is often off, it pronounces words weirdly.  Doing it for a full campaign and making it sound passable would be challenging.


Edit:  So I thought I'd have a go at this and see how good it was.  I had ChatGPT put together a command briefing for a new danger and put some retail Command Briefings into ElevenLabs on the cheap tier.  The attached is what it came out with.  I have to say it's a lot better than I thought it would be, especially on a first run.  It's not perfect, but it's firmly in the "good enough" tier.  If the pronunciation of "Nebula" sounds a bit weird it's because ChatGPT chose to say "Nebulae" for some reason"

[0rph3u5]

I enjoyed reading this.

To be fair the rest of us sounded like some grumpy elders right there (myself included)

Morally and legally I don't see it as fundamentally different from what we've already been doing.
[...]
I don't see why voice work is meaningfully different.  It's not like we're going to pay any of the original VAs to do it if we can't do it with AI, nor pay anyone else.  There's no material loss.

I think you are drawing a false equivalency here.

An actor's/talent's voice or likeness are not a comparable piece of intellectual property as they have an existence divorced from the fiction context they are embedded in. An actor/talent continues to be a person even if they stepped out of the recording booth/off the stage/leave the set. And that personhood is inseparably linked to their voice (both in the literal and metaphorical sense) and likeness as identifiable external characteristics.

While we are talking about professional actors/talents there is a case to be made about the commodification of this aspect of their personhood being consensual. However making that argument in full also means to acknowledge that said consent is conditional on remuneration. No payment, no consent, no commodity to be bought or sold.

Additionally, it would negligent not to consider that even with professional actors who take pay-check jobs as their livelihood, the pattern of payment = performance is not that simple.
There a good number of performances that only exist because circumstances align: the right director, the right script, the right moment in time, the right stage in their career etc. etc. pp. As a result some performances would not exist to be used as training data but for a house of cards of circumstance which made the actor/talent consent.



Furthermore, there is another point were your argument is lacking.
Your argument rests on the fact that the activity of this community, creating derivative works by reusing story concepts, 3D models, music etc. has been non-consensual from the start.

But it has not. Take a look at this section of the FS2 EULA (link to the wiki):

You [the owner of a copy of FS2] are granted a revocable, nonassignable limited license to create derivative works of this Software solely for your own personal noncommercial home entertainment use and may publicly display such derivative works to the extent specifically authorized by Interplay in writing. [Details on how to look up the specifics of that policy at the date of original printing follow]

By owning a copy of the game, each community member that owns the game, has been granted conditional consent to create derivate works and "display" them (i.e. share them). This is further reinforced by the fact that FRED exists in the retail versions of the games and its use is encouraged by providing a tutorial.

Said consent is conditional, but the condition is simply to that no money is being made from the derivative works as commodeties. Which is a condition easy enough to meet if you have desire for it.

[Grizzly]

There's a big difference between creating a cover for a beach boys song and creating a simulated version of the beach boys to make songs.

[Colonol Dekker]

This sums it up.

[mjn.mixael]

In popular media we see characters recast all the time and it's no big deal. I'd much prefer if someone really wanted Admiral Petrarch to be voiced in their new FS game that they just find a new voice artist. A good one could even try to approximate the voice well enough (see Red Skull in Avengers Endgame).

[Kie99]

I enjoyed reading this.

To be fair the rest of us sounded like some grumpy elders right there (myself included)

Morally and legally I don't see it as fundamentally different from what we've already been doing.
[...]
I don't see why voice work is meaningfully different.  It's not like we're going to pay any of the original VAs to do it if we can't do it with AI, nor pay anyone else.  There's no material loss.

I think you are drawing a false equivalency here.

An actor's/talent's voice or likeness are not a comparable piece of intellectual property as they have an existence divorced from the fiction context they are embedded in. An actor/talent continues to be a person even if they stepped out of the recording booth/off the stage/leave the set. And that personhood is inseparably linked to their voice (both in the literal and metaphorical sense) and likeness as identifiable external characteristics.

While we are talking about professional actors/talents there is a case to be made about the commodification of this aspect of their personhood being consensual. However making that argument in full also means to acknowledge that said consent is conditional on remuneration. No payment, no consent, no commodity to be bought or sold.

Additionally, it would negligent not to consider that even with professional actors who take pay-check jobs as their livelihood, the pattern of payment = performance is not that simple.
There a good number of performances that only exist because circumstances align: the right director, the right script, the right moment in time, the right stage in their career etc. etc. pp. As a result some performances would not exist to be used as training data but for a house of cards of circumstance which made the actor/talent consent.



Furthermore, there is another point were your argument is lacking.
Your argument rests on the fact that the activity of this community, creating derivative works by reusing story concepts, 3D models, music etc. has been non-consensual from the start.

But it has not. Take a look at this section of the FS2 EULA (link to the wiki):

You [the owner of a copy of FS2] are granted a revocable, nonassignable limited license to create derivative works of this Software solely for your own personal noncommercial home entertainment use and may publicly display such derivative works to the extent specifically authorized by Interplay in writing. [Details on how to look up the specifics of that policy at the date of original printing follow]

By owning a copy of the game, each community member that owns the game, has been granted conditional consent to create derivate works and "display" them (i.e. share them). This is further reinforced by the fact that FRED exists in the retail versions of the games and its use is encouraged by providing a tutorial.

Said consent is conditional, but the condition is simply to that no money is being made from the derivative works as commodeties. Which is a condition easy enough to meet if you have desire for it.

Likenesses that are inseparably linked to people are frequently used in fan works.  If someone's drawing art about Star Wars they might use Luke Hamill or Harrison Ford's face.  If someone were selling that art professionally they'd have to pay the individuals involved or whatever rights holders.

Does "the Software" encompass everything in the game?  You seem to be suggesting it covers story, animations, models etc.  Because any voices we made would be a derivative work from what was in the game, for non-commercial use.

I've got a vague memory of a mission someone did ages ago, I think it was a version of the opening FreeSpace cutscene with Lt. Ash.  Whoever made it had taken dialogue from the game and cut it up so the characters were saying different things.  It was nothing massively ambitious, it might have been editing together a few lines from Command to create new lines.  Supposing it existed as I remember it, should whoever made it have contacted the voice actor for Command and offered them money?  How is it different from what's done with an AI model except in terms of sophistication?

What if a person did a dead on impersonation of Robert Loggia, would that be permissible in a campaign?

There was someone on here creating 3D models of FreeSpace ships and selling them (I think at cost price, but they were still selling them).  Nobody had a problem with that.  That's not authorised in the EULA.

This isn't to say there aren't a lot of legitimate concerns about using people's voices.  It's going to have a devastating effect on jobs in the voice acting industry, there's a big risk of fake news, voice recordings are going to be rendered completely unreliable.  I just don't see their use in a free fan campaign that's obviously fictional and that a few hundred people are going to play for nothing as anything to be concerned about.

[mjn.mixael]

There's a big difference between chopping up existing files and recreating the voice. Voice artists and actors all over have come out strongly against AI being used to generate their performances without permission. You can find dozens of articles like this one with a simple google search. Are Robert Loggia or his estate okay with you recreating his voice? If you don't have permission, it's morally questionable and ethically low.

There's almost 8 billion people living on this planet. Surely it'd be easier, better quality, and more above board to get one of them to bring your new Admiral Petrarch to life.

[Mongoose]

My two cents: at the end of the day, I consume/play creative works because they are just that, created. They are a way of sharing in the skills and talents of another person, whether it's their ability to craft fantasy spacecraft out of nothing but Blender and their own imagination, to craft a piece of music that they devised, or to deliver a great vocal performance. That creative spark is a fundamental part of the human existence as far as I'm concerned. If the mass-market entertainment industry starts going the way of soulless algorithm-generated pablum, then they've officially lost me as a customer. I'll stick to what's being made by actual people.

And that's before you get into the massive ethical issues raised by applying generative AI to a deceased person's voice in order to generate a novel performance without their or their family members' permission. No thank you.