Support NeoGAF

[IbizaPocholo]

Sony maps out how first-party PlayStation studios are utilising AI tools during development

AI tools "allow our teams to spend less time on manual, high-effort tasks and reinvest time into building richer worlds and gameplay," says SIE CEO

www.gamesindustry.biz

Sony president and CEO Hiroki Totoki said that "AI will make it easier to take on more innovative and ambitious projects that were previously difficult to pursue due to constraints of cost and time."

"A core principle that guides our thinking about AI is that human creativity must remain at the centre. AI is a powerful tool – but it is not a replacement for artists or creators. It is an amplifier of human imagination and a catalyst for new possibilities."

Sony Interactive Entertainment CEO and president Hideaki Nishino stated that teams at Naughty Dog, San Diego Studio, and others have adopted Mockingbird, a tool generating facial models from performance capture.

This technology was used in recent releases like Horizon Zero Dawn Remastered.

"Importantly, we are not replacing human performers, but rather optimising how we process the data from these live captures. With Mockingbird, animation work that would have taken hours can now be completed in a fraction of a second."

Sony's first-party studios also use an AI tool for animating hair, which Nishino described as labour-intensive due to the large number of strands required.

"Our teams have accelerated this process by taking videos of real hairstyles and having an AI tool output a 3D model with hundreds of strands modelled."

Nishino emphasised that AI supports a more efficient production environment and improves discovery, helping games reach the right audience.

"These practical applications allow our teams to spend less time on manual, high-effort tasks and to instead reinvest their time into building richer worlds and gameplay for our players."

He clarified: "The vision, the design, and the emotional impact of our games will always come from the talent of our studios and performers," he said. "AI is meant to augment their capabilities, not to replace them."

Nishino also highlighted the use of AI in Gran Turismo, specifically in its AI-powered racing agent, Sophy. "Taking this further, our world-class creatives have already shown the ability to create amazing prototypes where NPCs with their own personalities can create a living, dynamic world for the players to explore."

He referenced the updated PSSR feature on PlayStation 5 Pro, and how it "uses machine learning to enhance image quality, delivering 4K visuals at high frame rates."

"Through our investments in AI and machine learning, we will continue to push the fidelity frontier forward," Nishino concluded. "We believe AI will unleash the creativity of our studios, power a more curated platform, and enhance the PlayStation experience for both players and creators.

"With our global player base, deep library of IP, and integrated ecosystem, AI is a powerful tool for us to deliver a truly cutting-edge entertainment experience."

 

[The Cockatrice]

>ctrl-f "ai will speed production allowing us to release more higher quality games over the years instead of once every decade"
>nothing found

k so its useless. gotcha.

 

[Three]

>ctrl-f "ai will speed production allowing us to release more higher quality games over the years instead of once every decade"
>nothing found

k so its useless. gotcha.

The stuff they listed is better than what you're suggesting. Sophy, and PSSR2 are far from useless. Don't know what that hair thing is though, first I'm hearing of it.

 

[Pejo]

They going to update this and re-release it every time they 'find' a new use for AI? Doubt it.

Give it 6 months and they'll be full-on

 

[Radical_3d]

Haven listed…

 

[RoboFu]

In related news Hulst says they now have unlimited ideas on new Horizon games.

 

[deriks]

Sugarcoating on something that gamers and devs don't like

To be fair, AI can really help here and there. The problem is that they like to extend those kind of things, and that's bullshit

Vote with your wallet

 

[bitbydeath]

>ctrl-f "ai will speed production allowing us to release more higher quality games over the years instead of once every decade"
>nothing found

k so its useless. gotcha.

These practical applications allow our teams to spend less time on manual, high-effort tasks

 

[BlownUpRich]

From the 1st picture:

• "Through our investments in Al and machine learning, we will continue to push the fidelity frontier forward."

Yeah AI will vastly help with increasing visual quality, Neural Rendering is where most of the "leap" will come from. Let's just hope they upgrade their game engines in order to advance to that goal where they'll be able to take small neural networks with all the parameters that their game engines can provide, in order for the AI inference model to accurately predict and calculate how things should look but at a vastly higher graphical quality. We will easily reach CGI film levels of quality with that!

 

[Angry_Megalodon]

Yeah, Horizon is an example of great storytelling, I can already see the benefits of this AI hair shampoo.

 

[squallbr]

It gonna reduce the loss at Bungie???

 

[Pedro Motta]

 

[pqueue]

Sonyslop

 

[The Cockatrice]

reply/quote won't work for some reason, but

bitbydeath

actual translation of your quote: we can fire more people now.

 

[archangelmorph]

The stuff they listed is better than what you're suggesting. Sophy, and PSSR2 are far from useless. Don't know what that hair thing is though, first I'm hearing of it.

AI wavy hair is better than getting games out the door faster?

Are you for real?

 

[Three]

AI wavy hair is better than getting games out the door faster?

Are you for real?

As I said, I don't know about wavy hair (what is it even doing? is there a paper on that somewhere). PSSR2 and Sophy are great though.

AI wouldn't get games out the door faster. It would just reduce headcount and get games out in the same timeframe with poorer quality. I'm not sure increased output would even help the industry, there are a lot of game releases as it is and things often get buried.

 

[cafecat]

Use ai in some work-flow not creative process

• facial/hair animation system driven by AI from capture data
• recommendation system and payment transactions powered by AI
• PSSR and other ML-driven graphics technologies to improve game visuals and give developers more tools on optimization.
• AI-powered NPC (rival)

Google survey, 90% game developers use AI powered tool in their workflow, the core is to restrict it in the repeated and complicated tasks, and left the developers more time in gameplay, storytelling, world and level design.

Some studios openly said that they regard AI as the tool to reduce their gap with these Japanese and western studios.

 

[IDKFA]

AI wouldn't get games out the door faster. It would just reduce headcount and get games out in the same timeframe with poorer quality.

Can you elaborate on this? I know there might be job losses, but why wouldn't it increase development time?

I'm not sure increased output would even help the industry, there are a lot of game releases as it is and things often get buried.

Game production now takes so long that a developer might be lucky to get one or two games out per generation. Considering the cost of development is increasing as well, taking so long to work on one game isn't sustainable.

 

[StereoVsn]

Devs: If you thought games were unoptimized as f, and only made palatable through upscaling and frame gen, hold on to your butts for AI produced code!

 

[foamdino]

The stuff they listed is better than what you're suggesting. Sophy, and PSSR2 are far from useless. Don't know what that hair thing is though, first I'm hearing of it.

That's a classic - take input and get an automated process to convert it into a format we want - the only change is using AI to do this instead of some tech artist having to take a reference and then go through n tools to get the desired output. I have no issues with devs using AI to do drudgery for them - this is *exactly* what AI is great at now. "Here is some data put it into this other format, write python or whatever, I don't care just give me the data in the correct format" - the only difference here is the input is a photograph and the output seems to be some "hair-strip model thing" that they want to use as an asset.

 

[Angry_Megalodon]

Some studios openly said that they regard AI as the tool to reduce their gap with these Japanese and western studios.

I have a better way. It's kind of revolutionary. Here it is:

how about hiring competent people who don't need 4 years just for pre-production and then cancel the project altogether and start over?

How about not needing 400 people working on a barebones game like Marathon?

AI is not going to fix any of this.

 

[riko]

Any developer not using AI at this point is going to be left behind. This ship has sailed.

 

[archangelmorph]

As I said, I don't know about wavy hair (what is it even doing? is there a paper on that somewhere). PSSR2 and Sophy are great though.

AI wouldn't get games out the door faster. It would just reduce headcount and get games out in the same timeframe with poorer quality. I'm not sure increased output would even help the industry, there are a lot of game releases as it is and things often get buried.

There are lots of problems with current games production cycles.

E.g.
- Talent ages out faster than you can ship a trilogy (or burns out and leaves to go make indie games)
- Product risk is sky high, meaning creative experimentation takes a back seat to "EVERY FUCKING STUDIO w/ PEDIGREE MUST GAAS!!"
- Sunk costs of bad decisions early in production make it harder to pivot and change course later as you've spent literally half a decade layering in content and code committed to the previous efforts that came before
- Etc etc

Of course AI can help get games out faster.

With the right tooling, production companies can shave years of man-hours off iteration work:

ART
- Concept artist use AI to rapidly assemble reference work, consolidate and align on themes & concepts to go into early sketches
- Artist start with by-hand low-fi sketches and then use AI to generate hi-fi pre-vis versions, review and tailor rapidly to align on initial design drafts
- Once design is locked they can either by-hand tweak or used tailored AI rework to get to the final designs
- 3D models AI-gen'd, tweaked and topology cleaned, bones, skinning, weights all gen'd & then by-hand tweaked
- End-to-end process is now 2-4 weeks (AI-assisted) vs 4-6 months (w/ mass off-shore outsourcing + onshore clean up)

CONTENT
- AI gen all "white box" NPC voice work, scripts, animations etc, to help designers get into higher fidelity pre-vis way way earlier. This means gameplay & scenario designs can be tested earlier iterated faster and polished faster one they're fully signed off. Much much quicker to get to "Is this cool/fun?"
- AI agentic QA; armies of QA bots should be used for the simplest of QA testing, noddy regressions can be caught super quickly and remedied early. Even more sophisticated AI agentic teams can be used to plan and test gameplay section walkthroughs, verify against progression breaks & showstoppers, etc

CODE
- Claude. Not much else to say here. This is the easiest area to see huge wins in code turnaround. Fully AI-assisted; no dev replacement needed

AUDIO
- AI-gen basic sound effects, no low-fi placeholder effects ever need to be generated by hand or sourced from commercial or royalty free databases. They can be tailored more closely to final SFXs to better aide designers to test immersion and verify quality earlier and faster

Basically, so much of games development is about experimentation. There is no world where AI tools and solutions don't help every aspect of this to allow teams to iterate faster because the cost of change is now far lower than it has been.Then once changes are locked then AI-assistance can help quality-mastering and really drive production values up higher than before (especially for smaller teams without e.g. world-class talented art teams).

 

[dgrdsv]

Sonyslop

SlopStation

 

[SHAKEZ]

I'll be real, AI can't be worse than the trash Sony is already putting out there...

 

[splattered]

Sonyslop

SlopStation

 

[splattered]

SlopStation

Thats what I get for not reading the entire thread haha

 

[GrayChild]

SlopStation

a.k.a. Sony's first party output after 2010.

 

[Three]

Can you elaborate on this? I know there might be job losses, but why wouldn't it increase development time?

You mean decrease? I'm saying there would be little change because it would be used to cut costs(headcount) rather than speed things up with the same amount of people.

Game production now takes so long that a developer might be lucky to get one or two games out per generation. Considering the cost of development is increasing as well, taking so long to work on one game isn't sustainable.

That's true if they maintain their workforce and speed things up with AI, output would be faster but I also think it becomes less sustainable when you increase output too. The industry doesn't have a lack of content/releases problem. There are a lot of games being released in the market already, to the point where even those that took years and years to make get buried rather quickly by some other release. Publishers would rather schedule and cut costs rather than increase output with slop.

 

[Moondoggy]

Gotta get all that pesky actual game developement out of their way so they can focus on what matters most, pushing the agenda.

 

[viveks86]

I have a better way. It's kind of revolutionary. Here it is:

how about hiring competent people who don't need 4 years just for pre-production and then cancel the project altogether and start over?

How about not needing 400 people working on a barebones game like Marathon?

AI is not going to fix any of this.

 

[The_hunter]

Automatic hair strand modelling, fixing 3d capture data.

Those seem like good use cases.

 

[ReyBrujo]

Sugarcoating on something that gamers and devs don't like

Devs do love AI, though.

 

[Physiognomonics]

I think there is a big difference betwen AI upscaling and generative AI. AI upscaling is not creating anything, it's just a tool to improve image quality.

 

[Mr Moose]

1 and 3 are good, not sure wtf 2 hair thing is about.

 

[DR3AM]

Imagine if they used GEN AI to make every office/room in Spiderman 3 unique. GAME CHANGER! I love AI.

 

[kpopSuperstar]

There are lots of problems with current games production cycles.

E.g.
- Talent ages out faster than you can ship a trilogy (or burns out and leaves to go make indie games)
- Product risk is sky high, meaning creative experimentation takes a back seat to "EVERY FUCKING STUDIO w/ PEDIGREE MUST GAAS!!"
- Sunk costs of bad decisions early in production make it harder to pivot and change course later as you've spent literally half a decade layering in content and code committed to the previous efforts that came before
- Etc etc

Of course AI can help get games out faster.

With the right tooling, production companies can shave years of man-hours off iteration work:

ART
- Concept artist use AI to rapidly assemble reference work, consolidate and align on themes & concepts to go into early sketches
- Artist start with by-hand low-fi sketches and then use AI to generate hi-fi pre-vis versions, review and tailor rapidly to align on initial design drafts
- Once design is locked they can either by-hand tweak or used tailored AI rework to get to the final designs
- 3D models AI-gen'd, tweaked and topology cleaned, bones, skinning, weights all gen'd & then by-hand tweaked
- End-to-end process is now 2-4 weeks (AI-assisted) vs 4-6 months (w/ mass off-shore outsourcing + onshore clean up)

CONTENT
- AI gen all "white box" NPC voice work, scripts, animations etc, to help designers get into higher fidelity pre-vis way way earlier. This means gameplay & scenario designs can be tested earlier iterated faster and polished faster one they're fully signed off. Much much quicker to get to "Is this cool/fun?"
- AI agentic QA; armies of QA bots should be used for the simplest of QA testing, noddy regressions can be caught super quickly and remedied early. Even more sophisticated AI agentic teams can be used to plan and test gameplay section walkthroughs, verify against progression breaks & showstoppers, etc

CODE
- Claude. Not much else to say here. This is the easiest area to see huge wins in code turnaround. Fully AI-assisted; no dev replacement needed

AUDIO
- AI-gen basic sound effects, no low-fi placeholder effects ever need to be generated by hand or sourced from commercial or royalty free databases. They can be tailored more closely to final SFXs to better aide designers to test immersion and verify quality earlier and faster

Basically, so much of games development is about experimentation. There is no world where AI tools and solutions don't help every aspect of this to allow teams to iterate faster because the cost of change is now far lower than it has been.Then once changes are locked then AI-assistance can help quality-mastering and really drive production values up higher than before (especially for smaller teams without e.g. world-class talented art teams).

What do you mean talent ages out before they finish a trilogy? it doesn't take 15 years to finish a game.