Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
×
AI Entertainment Games

Video Game Performers Will Go On Strike Over AI Concerns (apnews.com) 53

An anonymous reader quotes a report from the Associated Press: Hollywood's video game performers voted to go on strike Thursday, throwing part of the entertainment industry into another work stoppage after talks for a new contract with major game studios broke down over artificial intelligence protections. The strike -- the second for video game voice actors and motion capture performers under the Screen Actors Guild-American Federation of Television and Radio Artists -- will begin at 12:01 a.m. Friday. The move comes after nearly two years of negotiations with gaming giants, including divisions of Activision, Warner Bros. and Walt Disney Co., over a new interactive media agreement.

SAG-AFTRA negotiators say gains have been made over wages and job safety in the video game contract, but that the studios will not make a deal over the regulation of generative AI. Without guardrails, game companies could train AI to replicate an actor's voice, or create a digital replica of their likeness without consent or fair compensation, the union said. Fran Drescher, the union's president, said in a prepared statement that members would not approve a contract that would allow companies to "abuse AI." "Enough is enough. When these companies get serious about offering an agreement our members can live -- and work -- with, we will be here, ready to negotiate," Drescher said. [...]

The last interactive contract, which expired November 2022, did not provide protections around AI but secured a bonus compensation structure for voice actors and performance capture artists after an 11-month strike that began October 2016. That work stoppage marked the first major labor action from SAG-AFTRA following the merger of Hollywood's two largest actors unions in 2012. The video game agreement covers more than 2,500 "off-camera (voiceover) performers, on-camera (motion capture, stunt) performers, stunt coordinators, singers, dancers, puppeteers, and background performers," according to the union. Amid the tense interactive negotiations, SAG-AFTRA created a separate contract in February that covered indie and lower-budget video game projects. The tiered-budget independent interactive media agreement contains some of the protections on AI that video game industry titans have rejected.
"Eighteen months of negotiations have shown us that our employers are not interested in fair, reasonable AI protections, but rather flagrant exploitation," said Interactive Media Agreement Negotiating Committee Chair Sarah Elmaleh. The studios have not commented.
This discussion has been archived. No new comments can be posted.

Video Game Performers Will Go On Strike Over AI Concerns

Comments Filter:
  • by turp182 ( 1020263 ) on Thursday July 25, 2024 @07:55PM (#64656222) Journal

    I've been playing with text-to-speech, looking to convert older books to audio (philosophy).

    I've only looked into the open source stuff, coqui-ai and Tortoise.

    coqui-ai is very good, and offers voice cloning (so I can make audio with my voice if I wanted to), and voice conversion (converting a spoken audio file to another voice). I haven't tried those features yet, but the text-to-speech features are imperfect, but very good.

    Long text is actually pretty hard for consistency. But for short video game audio it's way past acceptable (combine with language LLM, same voice across languages).

    Image generators have a hard time with style fidelity, this audio tech solves that for voice.

    https://github.com/coqui-ai/TT... [github.com]

    • And good luck to the entire industry when generative AI can output entire games as fast as people can type up the prompts. Same with the entire entertainment industry as a whole. Star Trek got it wrong, the future is walking onto a holodeck and saying: "Computer, nazi shooter, sci-fi themed." Not playing some instrument or painting in your quarters.

      OR.... we could just admit that "taking away the work of a man takes something away from the man," and not be constantly looking for something to fulfill ourse
      • I've been playing with a few tools that allow me to say "I'd like to watch a show based on... With actors like... Wardrobe like...." And we're still sorta far off and the code capable of generation is extremely slow and unoptimized, you can tell we're years from this reality, not decades.

        Someone soon will optimize neural networks for non-linear equations and we'll see a progress jump even larger than when OpenAI showed up.
    • by AmiMoJo ( 196126 )

      We haven't reached the point where AI voices sound natural yet, IMHO. You hear them in YouTube videos regularly now, and they always sound slightly robotic.

      • I'm not sure shitty clickfarming Youtube videos are the best indicator of the state of the art :-P

        According to the Huggingface leaderboards ( https://huggingface.co/spaces/... [huggingface.co] ), Elevenlabs leads by quite a bit. I've tried it and it is indeed very good. Random youtube video with an example in the first minute here: https://www.youtube.com/watch?... [youtube.com]

        • by AmiMoJo ( 196126 )

          That video is what I was talking about, it sounds robotic to me. The intonation isn't quite right, it's deep in uncanny valley territory.

          • Mm, interesting. I guess the real test would be a blinded comparison between real and AI voice lines.

            There is some evidence that most people already can't hear the difference very well:
            https://www.techopedia.com/can... [techopedia.com]

            Also, have you listened to the voices generated by things like Suno? The music of course adds some 'cover', but nevertheless I'd say the results are very convincing: https://www.youtube.com/watch?... [youtube.com]

            • by AmiMoJo ( 196126 )

              It certainly sounds a lot better when singing. Maybe generating music gives the AI some hints about the cadence and intonation that a simple text prompt doesn't.

              That said it sounds like recorded rap, not live rap. Like they recorded the lyrics and edited the song together in a DAW, not like someone is just rapping to a backing track in real-time. Of course that's how most of these songs are produced, so it sounds the same, sounds normal. Or rather the real human versions don't sound quite real either.

              I've l

              • It certainly sounds a lot better when singing. Maybe generating music gives the AI some hints about the cadence and intonation that a simple text prompt doesn't.

                I think part of it may just be the nature of the training sets used for TTS voices. I'm assuming that those generally contain mostly fairly neutral deliveries, whereas the training set for voices in music is going to contain almost exclusively very emotionally and/or tonally rich samples.

                That said it sounds like recorded rap, not live rap.

                This is moving the goalposts, though. For video games and other places where voice actors have gotten work until now whether it is a 'live' recording or just a recording is fully irrelevant.

                It wouldn't save me any effort, unless I wanted my work lost in a sea of AI generated pap.

                This is a popular argument, but I

                • by AmiMoJo ( 196126 )

                  What I mean about recorded vs live rap is that in the studio what you hear is the result of many takes and some heavy processing. Auto-tuned as well. All the nuance that a real singing voice has gets taken out of it, making it sound artificial, so it's easier to reproduce.

                  I don't think it's just the training data, I think there is much less there that can give it away.

                  For video games it depends if they are going for a "cinematic" feel, where the vocals are highly processed too, or realism.

                  You are right abou

                  • What I mean about recorded vs live rap is that in the studio what you hear is the result of many takes and some heavy processing. Auto-tuned as well. All the nuance that a real singing voice has gets taken out of it, making it sound artificial, so it's easier to reproduce.

                    Ah, I see what you mean now. A friend of mine once gave me the analogy of photoshopped/airbrushed pictures of models when we were discussing the difference between 'authentic' and modern overproduced (pop) music. It made it easy to grasp how one might prefer the 'flawed'/'imperfect' version of things.

                    I do think it also serves as a reminder that most people don't care all that much about things sounding/tasting/looking artificial (or even just mediocre), however. As long as the stuff is cheap and tickles the

  • What is a video game performer?
    • Go play BG3...
      • by quall ( 1441799 )

        I have no idea either.... So just 1 game? Wow the industry will suffer greatly over this one.

        I Imagine that they'll either hire people who aren't "striking" or this will just accelerate their replacement.

        • Every single video game that has voice actors, motion capture, or even music. There are small names out there, and big names well known even outside of the industry. From GladOS, to Tiny Tiny, to War Never Changes, to Foul Tarnished! the voice actors provide a recognizable effect on their games.

    • I understand now. Gamers that record themselves playing games and stream it as a side business. Thxs
  • Risky proposition (Score:5, Insightful)

    by az-saguaro ( 1231754 ) on Thursday July 25, 2024 @08:21PM (#64656256)

    This is an interesting scenario where the "Law of Unintended Consequences" is a risk.

    Assume for a moment that the story is not about video game performers, but about the VG writers and programmers. If the industry then said "so what, don't need ya" and tried to develop games with AI, it would be an ultimate real world test. The upside risk for the grunts is that the AI experiment fails, the studios can't develop properties anybody wants, the VG consuming public complains about the lack of new titles or bastardization of existing games, etc. The studios learn that AI is a not-ready-for-prime-time money loser, so they go to the bargaining table with the developers, and all works out in the end, favorably for the VG programmers. The downside risk is that the real-person-independent AI-created games actually work and appeal to people and become a money-making market. The programmers who struck lose, and the industry is changed for ever, "screw you writers and your union". If the programmers did strike, it would be a fascinating scenario, better than the best of games, and I think it would work out in favor of the programmers.

    But, it is the video game performers who are striking. Like it or not, if the story and programs and action and cgi special effects, etc. are already worked out to satisfaction, then I can see where AI to generate the actors action and speech could work acceptably well. Personally, I think it is wrong and abusive to real people and talent, but I can see where it would make sense to studio executives precisely because it could be technically successful. If you are talking about a movie and replacing live action cinema characters with AI generated sprites, I can't see that working as well or the public accepting it. The studios would ultimately lose. But VG's are cgi to begin with, so replacing a meshed and skinned model with an AI generated model - it still looks like VG cgi, which is okay because it's a video game. If the gaming public accepts it, the executives can dump the real actors forever.

    I think that striking is a risky move for them, but I suppose they have little other choice. It will be interesting to see what happens. Maybe someone can hand them the Sword of a Thousand Truths (from the South Park episode "Make Love, Not Warcraft").

    • ha ha..... The Sword... ...but if you cut production cost you just have a fatter profit margin... the invisible hand will make it happen one way or another. Worst case, the companies just switch to AI while the actors are out on strike. Thats how Jazz was killed when recording became cheap and easy to do. The musicians went on a years long strike and they were replaced, long story short. Another scenario: small studios suddenly have a big boost in competitive advantage to be able to compete on a more even
    • If you are talking about a movie and replacing live action cinema characters with AI generated sprites, I can't see that working as well or the public accepting it.

      Not with today's technology but we are already close to crossing the uncanny valley - Rogue One's Tarkin and Leia were close and that was 8 years ago. Give it another decade or two and I suspect we'll be able to create CGI characters from scratch that look and act good enough that we can't tell the difference. This is not wrong nor abusive since, armed with such technology a lot more people will be able to tell their stories in ways that today require a major production budget and thousands of people, much

  • by khchung ( 462899 ) on Thursday July 25, 2024 @08:43PM (#64656292) Journal

    This will go the same way as any previous attempt to prevent machines doing human's job. It might work for while, but it would make the companies hiring them less competitive, and if they achieved their goal for the entire industry in the country, the whole industry sinks as competitors from other countries outcompete theirs, taking away not only the jobs that would have been replaced by automation, but also other upstream/downstream jobs that supported/relied on that industry.

    Yes, it sucks to be the ones replaced, but the rest of the world moves on with or without you. You either adapt or you get left behind.

    • This will go the same way as any previous attempt to prevent machines doing human's job.

      In the short term, I have no doubt you are correct.

      This time though the replacement product is (currently) inferior to the human product. There's a large difference between automating a fixed process and replacing creativity. The real problem with this - today - is that we're replacing creatives with computer models which are recombinants, not creatives. The results aren't anything new; they're mixtures of things we've already experienced. I grant that the majority of actors - both voice and stage - g

      • by khchung ( 462899 )

        This time though the replacement product is (currently) inferior to the human product. There's a large difference between automating a fixed process and replacing creativity.

        This is the oft used argument against machines replacing humans --- "but humans have creativity!" Yeah, if true, what are people worried about?

        Unfortunately, first, just look at what creativity there is in the supposedly most creative industry, the movie industry. How many rehashes, sequels, prequels, basically more of the same staff gets released every year, vs real creative movies? Why worry about AI generating the more-of-the-same stuff, while humans continue to create actually creative movies?

        Second,

    • by eepok ( 545733 )

      You cannot stop the sun from rising, but you can wear a hat and sunscreen.

      They're not trying to abolish technology. They're trying to limit it's affect on their jobs. That's not only possible, but normal.

  • These unions make up stuff based on a lack of understanding of other stuff.

    Then they harass other people with their stuff.

    Enough of saf-astra and other idiots trying to get a pound of flesh for having done nothing more than they normally do.

    AI training doesn't steal intellectual property.
    None of this is piracy.
    If you make your content available for people to view, some will view it unaided, some will wear contact leanses, some will wear glasses, some will use augmented reality, some will use search engines,

    • The issue is not that LLMs will view their work. It's that the LLMs will then be used to generate new images, animations, or sounds, in their likeness.

      This difference should be obvious, and is quite critical. They don't mind being seen and heard. They mind being duplicated and then disposed-of.

      Quite a reasonable grievance, I would say.

      • This difference should be obvious, and is quite critical. They don't mind being seen and heard. They mind being duplicated and then disposed-of.

        Why would they need to duplicate them? Turns out that you only need to be within spitting distance.

        https://arstechnica.com/tech-p... [arstechnica.com]

        What's interesting in this case is the target voice wasn't even that of Scarlet Johansson, rather it was a modulated version of her voice that they were targeting. What they got was basically indistinguishable, and they didn't even need Scarlet, all they needed was for somebody else to use her own normal speaking voice. Scarlet can't claim to be wronged here unless she believes

  • Existing AI is already good enough. Worst case can hire some random ugly non-union people to act out the parts and then face-swap in a generic "good looking" face/body using AI.

  • Yeah I understand they are afraid of their job, but if AI can be trained on 'youtube' videos to do motion, it is a lost cause. You can prohibit a company from using AI instead of actual performers, and nobody can force me to use SAG actors etc for my work. AI will replace a lot of jobs just like machines and robots replaced a lot of jobs in factories and software replaced a lot of jobs in the office. Society has to deal with people not having a job for earning money for housing, clothing, food and healthcar
    • I think you may underestimate black-balling the union if you are an entertainment company of any size. First of all SAG is just one of the many unions, IATSE, WGA, DGA etc. If you want to make stuff all by yourself or with scabs you can, but you won't be getting the best in the field by a long shot.
  • It seems the running gag with AI is that the content can't be protected because there was no "artist" to grant the rights to. So I wonder if going all AI would allow "fair use" of the properties that they can't copyright/trademark/etc.
  • In other news, buggy whip manufacturers have been on strike for over a 100 years over unfair practices impacting their businesses after widespread usage of cars began. Surely someday the market will cave in, just like for these actors.

  • At least how I see it. If gaming companies allow AI to infiltrate all aspects of game development with the ultimate goal to have AI build a game from scratch with just prompts, all they'd need are some prompt specialists. The outcome of this will be significantly lower costs for the gaming companies and massive unemployment for most people in the industry today. The flip side of this will be that gaming companies may innovate their way to extinction. Once you have those tools to make games from scratch, wh
  • by mrfaithful ( 1212510 ) on Friday July 26, 2024 @10:20AM (#64657466)

    VG voice acting was a bad idea. We went from tight, decently written dialogue text that you could page through quickly into this sprawling nonsense read in a slow drawl with plenty of pauses to account for animation delay. Just like FMV, it was worth a shot when it was new technology but it was best relegated to minimal usage once we established that it wasn't a net benefit to gameplay. Fully voiced visual novels or your "walking sims" can be the exception where the voice acting and cutscenes are the draw for the fans, but for gameplay focussed games, get 'em out.

    A story I can only half remember (was it the monkey island writer?) was that the reason CD based point & clicks had lesser dialogue was because beforehand he'd have been writing, re-writing and editing the text right up until gold master. But with voiced dialogue he had to have the script finalised long before the game itself was finalised and retakes were costly and time consuming so the weak lines and poor pacing had to stay. Even in games where the voice acting was done after the fact in a CD edition, it's still preferable to me to play the floppy version.

    • There are many games though where the voice acting is done very well and provides extra depth. That's why game companies do this, despite the expense. Sometimes they trim down the number of lines to keep the costs down, other times they get lower paid actors for the random about-town background voices (possibly using their own employees), or being only partially voiced, but cutting them out will hurt most games. Any action oriented video game doensn't work if it pauses to let the player read the in-comba

How many hardware guys does it take to change a light bulb? "Well the diagnostics say it's fine buddy, so it's a software problem."

Working...