Gran Turismo 7: Latest news and discussion thread

  • Thread starter Thread starter sems4arsenal
  • 50,903 comments
  • 7,889,935 views
My lucky shot will be a Motegi in around November :D

And Circuit Gilles Villeneuve isn't out of the window only because it wasn't shown on Montreal GTWS event ;) ;)

But after PD did break "american tracks" cycle. I could expect everything now. I wonder is Silverstone a possibility since it was made for GT6... would PD scrap that track after one entry?
No new track until 2025
 
Last edited:
What if Gilles Villeneuve was just the internal name PD gave Eiger in the files as they worked on it for its release at the Montreal event
 
Last edited:
What if?

I think that's a stretch given there were image files found too, weren't there?
Correct, it was found in 2 different places.
Scapes files originally, along with Albert Park, and more recently in the more familiar Time of Day settings list.
 
Last edited:
What if Gilles Villeneuve was just the internal name PD gave Eiger in the files as they worked on it for its release at the Montreal event
Maybe in that case they'd call it "track_montreal_gtws" or something like that? Think maybe Gilles Villeneuve would be too specific to the circuit to not be that, imo.
 
I always thought Gilles Villeneuve might have been Lake Luise since they are both in Canada.

EDIT: or not.
There was a generic test_snow_oval that likely became Lake Louise.

Speaking of old datamines reminds me - the really old track ID datamine, the one that came from GT Sport years ago. The one that had the 4 GT1 tracks and Road Atlanta (but not Watkins Glen). Does anyone recall if that had Eiger in it as well?
 
Does anyone recall if that had Eiger in it as well?
I believe Eiger was in one of the old datamines, yes.

Along with Tokyo R246, but that might have been scrapped 'cause of Kaz saying city tracks taking too long to make.
 
I believe Eiger was in one of the old datamines, yes.

Along with Tokyo R246, but that might have been scrapped 'cause of Kaz saying city tracks taking too long to make.
Maybe one they will make or something. He sometimes change his mind
 
In 2022 we did get 2 tracks. Watkins Glen amd Road Atlanta
In 2023 we did get 2 tracks. Grand Valley and Lake Louise
Why would 2024 be so special? Some people thought we won’t get any new track at all and here we are :)
See you in 2025, when PD doesn't adds a new track in 2024
 
Nearly screamed when I saw the new RUF thinking it was an 992 Turbo S, but a new RUF is always welcome. Never though I'd see the Gallardo and Scuderia in a Gran Turismo game ever again, hopefully we can also get the pre-facelift version with that glorious 5.0 V10 sound :lol:

I'm happy to see Eiger again, always had a soft spot for montainous circuits. Here's hoping they can bring back Matterhorn and Cape Ring too.

View attachment 1370895
Matterhorn was the track that made GT6 stand out for me. Now that there’s precedent for Swiss tracks with mountainous backdrops making a comeback on current gen, I would be over the moon to see it return
 
Maybe one they will make or something. He sometimes change his mind
"Unless some kind of technical innovation happens." Those are his words from that interview.

So unless said tech innovation does indeed happen that'll speed up dev times, we shouldn't expect any.
 
"Unless some kind of technical innovation happens." Those are his words from that interview.

So unless said tech innovation does indeed happen that'll speed up dev times, we shouldn't expect any.
Perhaps high quality generative AI trained on their Tokyo Expressway building assets? With reference images they took from the real Route 246 for textures? And manual adjustments afterwards if needed. Only way I can see accurate city tracks coming with their level of detail. I know they have procedural city building in the works, but unless they can tune it to produce at least decently location-accurate buildings, I don’t think they’d brother since they are high on making the locations accurate. Which I do appreciate but obviously it does increase dev time.

That said, just did some research, seems AI tools (public ones at least) aren’t great at complex 3D models. Buildings are at least somewhat among the more “basic” shapes for the most part, so maybe their own AI tailored for cities could do the job. I’m sure they could get some great programmers given their budget, and if it’s not doable yet, maybe in 5 years or so? AI is improving at a crazy rate after all.
 
Last edited:
So how’s the new physics?
According with some spanish youtubers that I follow and are closer friends to some GTWS spanish competitors:

-Suspension changes (specially on dirt surfaces)
-Braking is now more sensisitve (hard braking will cause the car to lose stability)
-Tire degradation o avoid people on still running with tires on red

One of them was even talking some craziness about tire puncture happening if you continue driving with tires on red, but he said to take that with a grain of salt 🤭🤭🤭
 
Perhaps high quality generative AI trained on their Tokyo Expressway building assets? With reference images they took from the real Route 246 for textures? And manual adjustments afterwards if needed. Only way I can see accurate city tracks coming with their level of detail. I know they have procedural city building in the works, but unless they can tune it to produce at least decently location-accurate buildings, I don’t think they’d brother since they are high on making the locations accurate. Which I do appreciate but obviously it does increase dev time.

That said, just did some research, seems AI tools (public ones at least) aren’t great at complex 3D models. Buildings are at least somewhat among the more “basic” shapes for the most part, so maybe their own AI tailored for cities could do the job. I’m sure they could get some great programmers given their budget, and if it’s not doable yet, maybe in 5 years or so? AI is improving at a crazy rate after all.
AI could help. But seeing that it's so controversial right now (especially in the art/fanart space), it's probably not worth the negative PR. =/
 
AI could help. But seeing that it's so controversial right now (especially in the art/fanart space), it's probably not worth the negative PR. =/
If they train it on their own data, the ethical debate should be more kind, no? I totally support the artists who are upset with AI being trained on their works when they didn’t consent, but if they train it on studio assets, since they are owned by PD, that side shouldn’t be an issue.

Although I just realised that the individual artists at PD may not want their work used for AI in any way. Which I understand. Maybe hold a studio vote and if it’s unanimously in favour, go ahead with the plan.

It being private data moving towards a singular design goal for the game when it’s otherwise not feasible could maybe change someone’s view as long as it’s not then used in other, exploitative ways. But ultimately that decision of consent rests on the individual and it should be respected.
 
What about any bug fixes? and other minor things we should be seeing in the update?
 
hoping they add the ability to sell unwanted gift parts an engines 🙏🏽
Oh that yeah. That's always been weird because when checking out the parts and engines in storage, they're listed as having a "recommended retail price" (which is the same price it costs to buy), implying you can set your own price with that as a reference, and sell it somewhere.

That'd perhaps work a bit like the storefronts or auction house from the Forza series where you can sell things to other players, as an alternative to selling back to the game itself.
 
Last edited:
I was just re-watching the update trailer, but on the Manufacturers Cup live stream, as it's where you can hear the crowd cheering and reacting to it, just so I could listen carefully in which parts people reacted the loudest. There were four major parts:

-When they saw there was a new track and recognized Eiger.
-After the GT Sophy part was presented.
-When the Impreza Rally Car was shown.
-When the trailer ended.

Considering that Kaz was there, I wonder if these reactions somehow set the tone of what people are expecting right in front of him. 🤣
 
If they train it on their own data, the ethical debate should be more kind, no? I totally support the artists who are upset with AI being trained on their works when they didn’t consent, but if they train it on studio assets, since they are owned by PD, that side shouldn’t be an issue.

Although I just realised that the individual artists at PD may not want their work used for AI in any way. Which I understand. Maybe hold a studio vote and if it’s unanimously in favour, go ahead with the plan.

It being private data moving towards a singular design goal for the game when it’s otherwise not feasible could maybe change someone’s view as long as it’s not then used in other, exploitative ways. But ultimately that decision of consent rests on the individual and it should be respected.
Thing is, you cannot train on solely your own data. Generative AI only exists through being fed as much data as they can manage, and it still ends up looking quite terrible. To train on "your own data" means to feed it some of your stuff and tell it to make something like that, but it still has to use everything else it's eaten up to poop out something usable.

That, along with the insane energy cost of generative AI, makes it impossible to do ethically. Anyone who claims to have used AI trained only on their own data is either lying or misunderstanding how it works. You need that massive database for it to work, and no one company owns enough to manage that. Disney could feed an in-house AI model all of their movie scripts (including ones they've bought over time) and it would be unable to produce a somewhat workable script for a new movie. Even after editing out the glaring issues it would still be very obvious that it was just taking bits and pieces of their past work. Models like ChatGPT had to scrape as much of the internet as they possibly could to get it to become the best predictive text tool in the world. Still, it's a predictive text tool; it barely maintains coherence if you ask it to do more than answer a simple question to which it had the answer scraped from some website already.
 
Thing is, you cannot train on solely your own data. Generative AI only exists through being fed as much data as they can manage, and it still ends up looking quite terrible. To train on "your own data" means to feed it some of your stuff and tell it to make something like that, but it still has to use everything else it's eaten up to poop out something usable.

That, along with the insane energy cost of generative AI, makes it impossible to do ethically. Anyone who claims to have used AI trained only on their own data is either lying or misunderstanding how it works. You need that massive database for it to work, and no one company owns enough to manage that. Disney could feed an in-house AI model all of their movie scripts (including ones they've bought over time) and it would be unable to produce a somewhat workable script for a new movie. Even after editing out the glaring issues it would still be very obvious that it was just taking bits and pieces of their past work. Models like ChatGPT had to scrape as much of the internet as they possibly could to get it to become the best predictive text tool in the world. Still, it's a predictive text tool; it barely maintains coherence if you ask it to do more than answer a simple question to which it had the answer scraped from some website already.
Hmm, I never thought about that, that it would require a huge amount of data. And didn't think about the energy cost. Didn't really know it took so much energy but that makes total sense, with all the training they need to do, right? Maybe procedural is the way to go then, using their Tokyo Expressway assets as a base for it? Long as they have a good variety of buildings, maybe touch up any which need replacing for accuracy, might be a good middle ground. At this point I would rather "Tokyo R246-like et al." environments if it means we get city tracks. They're great variety, imo, and unless it's something like Monaco or a famous vista, something which gets the "likeness", but maybe isn't 100% accurate in environment, will be fine. Especially when the alternative is no city tracks at all (or very very rarely).

Appreciate the input!
 
🤔
CCGT on the track_0.jpg
 
Back