Gran Turismo Sport: General Discussion
- Thread starter Formidable
- 47,132 comments
- 4,757,387 views
- 14,435
- Los Angeles, CA
- Lma_robot
- lImaRobotl
Say they import them and they don't stick out like a sore thumb? I'm not sure why you'd go on arguing if you agree that no one has seen them together,as all it seems that you're doing is exactly what you where telling @Samus that he was doing. You are the one being hypocritical.
Then that will suck, but what if they are compatible? Still all guesses from both sides. Forza obviously had cars that where ported from last gen(some even have the exact same paintshop problems across both gens), and it didn't seem to stop the livery editor. I imagine PD would be able to think of something in that regard.
It's weird how you tell us how wrong our opinion is because we haven't seen it. I take it you've seen a damage model?
Highly ironic. Either way, your blind faith comment doesn't really fit here. I'm having faith that people will accept a gap because both parties are already accepting it. Those with the standards already accept gaps, and those for premiums are already accepting that they'd want them in GTS.
People that where for standards where not against premiums, so the insistence that this just wont work is odd. they'll be happy that cars, in general, get added to the game.
However, if its those people that get upset because their very special single car isn't in the game, be it premium or standard, then that's something that can't be helped.
Yes. All you did was post a bunch of guesses and hypotheticals to say someone is wrong.
Neomone
(Banned)
- 133
- Australia
Who saw the difference between GT5P premiums and GT6 premiums as a massive deal?
Of course whether people care or not is going to be subject to opinion. But the difference between GT4 standards and GT5P premiums was clearly visible whether you cared about it or not. The difference between GT5P premiums and GT6 premiums is difficult to impossible to spot outside of specific conditions that are really only relevant to a subset of players (photographers that like closeups). If you're further away than the LOD0 model then there's no functional difference.
Like I said, I can only judge what I've seen. What I've seen is that they're pretty close. If GTS starts showing significantly improved models then I'll change my tune, but I can only work with what I've got. It seems like a better idea than just assuming that they'll look bad on no evidence at all.
I think you're making a big fuss about the quality of the premiums. They're still higher poly count than the cars in games like pCARS, which while not up to the standards of GTS models are perfectly acceptable looking for the PS4. I would expect premiums to look even better. I see no reason to assume that they would look out of place without actually seeing it with my own eyes.
Again, I can only work with what I've got. If that were true, then yes that would be a massive gap. One could question why cars made at the end of the PS3 gen weren't compatible with features that Polyphony has been trying to include for a decade now, but it would certainly be a major division.
We're yet to see the damage model and we have limited footage of the livery editor. When we get more maybe this will become apparent.
- 8,715
Actually I hadn't considered the sound aspect recently.
I had already commented that the lower car count makes it easy to do the stop gap sound "solution" that we've been hearing.
They must have done a lot of work with all the cars already on their eventual target sound, no need to throw that away or duplicate it all several times over, either.
I don't think the "tessellation" method PD uses has anything to do with displacement maps, at least not expressly. I think they used that name because people already hear it. PD don't use full body textures as it is, they certainly won't do it for displacement mapping!
Video of it in motion looks like adaptive progressive meshes, with certain edges / groups selected for additional smoothing at higher LoDs beyond the number of stored vertices. That smoothing in itself is more like adaptive subdivision according to certain primitives, rather than the simple grid subdivision according to one primitive (the plane) that you get with displacement mapping.
The key difference is the primitives (defined per edge / group) determine the shape of the subdivision (mostly circles and arcs) in PD's method, whereas the map determines the shape of the already subdivided grid in "ordinary" displacement mapping. Of course, PD can store high-resolution polygonal data in a map (texture) and re-apply it in real time just the same, but it need not be planar in application, rather based on surface normal displacement - but they'd do that in a similar way to how they apply decals: only where the detail is necessary.
So the progressive mesh part is perfectly automatable (just rearrange the order vertices, edges and faces are stored), the edge detection can be automated with hand-deletion and tidying of that output and finally defining / morphing of primitives to match the edges found (splitting as necessary).
I had already commented that the lower car count makes it easy to do the stop gap sound "solution" that we've been hearing.
They must have done a lot of work with all the cars already on their eventual target sound, no need to throw that away or duplicate it all several times over, either.
I don't think the "tessellation" method PD uses has anything to do with displacement maps, at least not expressly. I think they used that name because people already hear it. PD don't use full body textures as it is, they certainly won't do it for displacement mapping!
Video of it in motion looks like adaptive progressive meshes, with certain edges / groups selected for additional smoothing at higher LoDs beyond the number of stored vertices. That smoothing in itself is more like adaptive subdivision according to certain primitives, rather than the simple grid subdivision according to one primitive (the plane) that you get with displacement mapping.
The key difference is the primitives (defined per edge / group) determine the shape of the subdivision (mostly circles and arcs) in PD's method, whereas the map determines the shape of the already subdivided grid in "ordinary" displacement mapping. Of course, PD can store high-resolution polygonal data in a map (texture) and re-apply it in real time just the same, but it need not be planar in application, rather based on surface normal displacement - but they'd do that in a similar way to how they apply decals: only where the detail is necessary.
So the progressive mesh part is perfectly automatable (just rearrange the order vertices, edges and faces are stored), the edge detection can be automated with hand-deletion and tidying of that output and finally defining / morphing of primitives to match the edges found (splitting as necessary).
- 489
- England
No I am not being hypocritical in the slightest, there is a gap in quality, you don't have to take my word for it though, the only people's opinion that matter on the subject have seen them, PD and the fact is, they have been cut from the game. Or is it lazy PD? Greedy PD? Have no clue what they're doing PD?
Come on now.
Old texture sets and materials is not a guess, it's a fact. Old LOD sets and methods is not a guess, fact. Lack of polygons up close is not a guess, you know, you're gonna be surprised to learn that that is a fact. 👍
3 facts there mate, 3 facts when combined or not make a gap, 3 more facts than most people.
Oh, it's like as if they already had a livery designer for FM or something wasn't it?
The only thing weird here is your lack of comprehension of the word hypothetical. I clearly stated the distinction between me hypothetically speaking and facts.
Belief without true understanding is the definition of the phrase 'blind faith' It fits perfectly, like a glove.
Same engine, same texture sets, same modelling method, not the same situation.
I could not agree more, but let's get back to the point, that's still a two tier system if included. I don't know how many times I have to point that out. You probably make some other point I'd like to address, might come back to address later.
It's actually a target mesh but not stored in the conventional fashion. Still from a high poly source and the output is supplied by something closer a map than any 3d type file (really cheap on DP compared). Hence my use of the term 'map'.
- 1,984
- Barcelona
- mannjonusa
We need to ask Kaz, Will Cizeta V16T, Countach 25th Anniversary, Dino 246GT and other new premiums in GT6 be in GTS ? if not, why the exclusion, it was newly built premium for GT6, and it was very detailed compared to GT5 premiums, it would be a waste to just drop it.
Those new gT6 premium would easily add the already paltry car list in GTS, if they are not included yet in the 130+ car list
Slightly related to this, (and prompted by a gander at the badass EAG tunes in your sig - damn that 300SL is sick), have they announced any vintage cars for GTS?
Personally, I'd love that. A whole section dedicated to old machines...
(Maybe also nascar? Some other random stuff?)
Last edited:
- 8,715
- 2,534
- Johannesburg
- Tayeedz1
I've written a script for a GTS video I'm planning.
Funny enough, this video was meant to be out last year. Eish when life takes over....
With all these events and tid bits coming out it's changed the script significantly.
I think I'll get to it next week maybe. Time to actually get it out there.
Funny enough, this video was meant to be out last year. Eish
when life takes over....With all these events and tid bits coming out it's changed the script significantly.
I think I'll get to it next week maybe. Time to actually get it out there.
Neomone
(Banned)
- 133
- Australia
If you say so. That seems awfully dismissive to me. It sounds a lot like you're saying "they were in another game, they're different by definition".
Which I suppose is true, but it doesn't get us very far. The question is more "will the average user be able to tell these things apart?"
The answer with GT5P premiums and GT6 premiums was "not really", and so they were considered a single tier despite being different. The answer with GT6 premiums and GTS premiums is unknown, but the visual evidence we have so far to my eye tends towards "probably not".
The thing is that even the premiums on PS3 were a multi-tier system in GT6. We had GT5P premiums, GT5 premiums, and GT6 premiums and all had different levels of quality and features.
But most people were OK treating them as a single whole because the differences were minor.
If the differences between PS3 premiums and GTS premiums are major, like with your example of potential damage and livery editor incompatibilities, then that definitely seems like a two tier system. But if the only differences are minor and only visible under close scrutiny, then what's the point of the division?
While I see how it could be a two tier system, I don't see anything so far that suggests that the PS3 premiums would be unable to be integrated smoothly in what would at least appear to your average user to be a single tier.
- 1,335
- WA
- TEAM-SWD-daus26
That one screenshot or video where the vision GT with no windows (peugeot?) looking out of place from the rest of the grid, what's the reason for that? I wish I'm good with finding screenshots. I just thought that car may need to be redone perhaps same way the rest of the premiums from GT6 may need, cause the car didn't look out of place in GT6, but do in GT Sport. To me at least.
- 8,187
- Kaikoura, NZ
- Nato777
Except for the fact that PD have decided not to use them but rebuild them completely for GTS - that in itself could suggest that the PS3 premiums are unable to be integrated at least without too much compromise. Did they do that just for the hell of it, because they can and maybe they pride themselves on having the best possible models on the platform? Possible I suppose, but there is more likely to be technical reasons due to any number of things already mentioned such as new rendering/lighting engine, livery editor, VR or the as yet mythical damage system.
Problem is we just don't know what the reasons are and unless someone specifically asks Kaz what the difference is from GT6 to GTS that meant the models need to be built specifically for it then it is likely to remain a mystery.
- 903
- Tasmania
- RobUrabus
I'm sure there was a properly technical discussion some months ago outlining the reasons for the re-design
I've got no idea about how these things work technically, but I believe it was something to do with the way the damage needed to work and show proper deformations in the surface.
Rather than an old school "solid-box" model that could only have small polygons removed to show damage.
I could be 100% wrong or it was part of another discussion.
- 8,187
- Kaikoura, NZ
- Nato777
Pretty sure that technical discussion was still only based on hypothetical ideas though. I think it's dubious at best at the moment to attribute a damage system that has yet to be seen as the reason for PS3 Premiums not being used in GT Sport.
- 489
- England
It's best to go back a post.
No other mesh exists besides the main mesh, polys go up and down on this base mesh via 'target maps'. These maps are 2D in looks only, they contain all the data for all 3 axis, you could actually build a replica of the original mesh via these maps as they contain all the data from the original in a very cheap form. No heavy data involved as all the data points are baked out.
I get what you are saying here but in reality would not be a practical solution on its own without a target map. At some point or some curve or corner to be precise, you are always going to need a target that is supplied in the form of a map or mesh for definition. Mesh is ruled out because of the obvious impact it would have on perf.
So the progressive mesh part is perfectly automatable (just rearrange the order vertices, edges and faces are stored), the edge detection can be automated with hand-deletion and tidying of that output and finally defining / morphing of primitives to match the edges found (splitting as necessary).
How would you capture and apply that data? Probably gonna need some target/displacement maps for that, they are tried and tested and the cheapest way after all. They give you the definition where needed or where not.
Whist everything you said can be true, the workflow involved is not a practical or accurate enough solution and sounds like more work than actually building a higher mesh. Also with this being hardware driven I'm pretty sure they'll be using the standard set by the hardware and the methods used by other developers, definitely won't be too far from those methods anyway.
To get back to the two birds one stone metaphor I used, if the topology is wrong to begin with for the use of AT (which is the most probable explanation) They would still have to start from the beginning of the modelling process. I don't know if that renders everything I said above pointless or not.
No, I'm saying they're different in design. Big difference.
The average user will never know, PD made that decision for them, either they looked at them and thought no by themselves or took advice from outside on board.
It seemed like a fair amount of people wanted consistency, all the people are now getting consistency. Turns out some people no longer want that now, despite claiming they did, the most important thing in fact I have read from them in the past, turns out it's no longer the most important thing any more. Also questioned others as to why they were content with the inconsistency. Very hypocritical to say the least these people.
The difference between those two games models where polygons, maybe texture quality and a software driven AT solution, the difference between these two games models are game rendering engine, more polygons, PBR materials and a hardware based AT solution, oh and plus under the hood the obvious perf gains for not using expensive LOD data points. Mind the gap
Like I said I don't disagree, I still wouldn't care that much if they made it but there would be a sizeable gap if they were straight up ported. PD must feel the same way and maybe they may return for another title where they have mixed and matched different era models before, who knows (slap some PBR mats on at least if you do PD).
That's far from the point I originally made though, my original point was that they would be inconsistent. Is there factual inconsistencies between the two sets? Yes. Did the poster I quoted say he wanted consistency.....
Many times he's stated that by his own admission! Not only that but it is....
The most important factor no less! Turns out it isn't the most important factor after all!
See where I'm going with that right there?
- 14,435
- Los Angeles, CA
- Lma_robot
- lImaRobotl
Youre being hypocritical because you keep going on about how "no one has seen this or that" but continue to keep making guesses about things we have not seen or has not happened. I never disagreed that there is a gap, more so that it just seems that gap is pretty negligible and that it in now way is similar to the gap between premium and standards. It resembles that of premiums starting with gt5 and those introduced in GT6. Both are already extremely detailed. They where fine as is, really. And im sure those for and against standards would have agreed with that, they where, and still are, the cream of the crop. There was no need to just drop them completely.No I am not being hypocritical in the slightest, there is a gap in quality, you don't have to take my word for it though, the only people's opinion that matter on the subject have seen them, PD and the fact is, they have been cut from the game. Or is it lazy PD? Greedy PD? Have no clue what they're doing PD?
Come on now
However, maybe it had less to do with the actual vehicles, and more so the streamlined, goal oriented focus of the game. To make it fall in line with the dedicated racing events it's holding. Still there was no need for such a drastic cut.
Yes those are obvious facts, none of which I was talking about. When I said you where guessing is when you say what if they aren't compatible.
What reason would it not be compatible if premiums where supposedly already modeled in pieces, rather than one uniform guestimates shape like the standards?
If that it being past gen is such a problem, why not start a livery editor based off the already amazingly detailed past gen models, because it doesn't seem like it's something that needs to be that advanced in the first place. Hell, Forza's is comprised of just moving and editing very basic shapes over the side of the vehicle that you're looking at, something that seems entirely possible to do with GT, regardless of the slight differences we've seen so far.
Livery editor or not, your saying that it can't happen because there is a generational gap and that they aren't modeled the same, which was the same thing that has supposedly happened with Forza.Oh, it's like as if they already had a livery designer for FM or something wasn't it?
Well, good thing is that youre not disagreeing that you're literally doing exactly the same thing you're telling others they shouldn't do because we haven't seen it. Glad to know that you realize it.
Like I said, ironic.
- 489
- England
- 137
- Japan
I said, it is technically possible. The premium car in GTS. But, I do not think that all the people are welcome. The same as the GT3 A-Spec. So Standard cars in GT5 has been prepared. As a result, the standard car by not a few people have been criticized. So PD there was a need to prepare the only super-premium.
Now GT6 premium cars and super-premium car is it does not look like there are so many gap, But after its gap are concerned that becomes a problem.
Now GT6 premium cars and super-premium car is it does not look like there are so many gap, But after its gap are concerned that becomes a problem.
- 357
- Singapore
- 2,789
- mount doom
- coilerdh
- no.no.
So much work that PS4 cant even handle in real time- that will be scraped for next gen
- 16,143
- Melbourne
- ScottPuss20
- CheetahsMeow
This thread... Ugh, it's like a warzone.
- 28,470
- Windsor, Ontario, Canada
- Johnnypenso
- 357
- Singapore
- 50,926
- Australia
- SpacedustDaddy
- 8,715
The mesh itself is most likely stored as a hierarchy, or "tree". See here. It requires a little, but not much, more data than the original high-LoD polygon list, just stored in an order that makes pulling it out progressively extremely easy. No need for discrete LoDs etc. and much higher LoD range than displacement mapping in practice. If that's what you mean by "map", then all's good. The tessellation (subdivision) is performed in addition to that adaptive, progressive detail streaming of the existing stored polygons.
You can't "bake out" data, unless you're talking about some kind of lossless compression scheme. Information is information, if you want detail, you need information; you need the data. What you're describing still sounds like normal-based displacement mapping - these have a fixed spatial resolution, so artifacts and / or loss of definition would eventually be visible as you zoom in. But such artifacts do not exist on PD's models, except where tessellation is not applied. (The storage saving comes from the spatial information being implicit: 1D with reference to calculated surface normals instead of stored on a 3D grid, and only because that calculation can be afforded in exchange for freeing up some precious memory bandwidth at run time).
I was thinking more along the lines of a hybrid parametric and mesh approach.
The conversion to progressive mesh is trivial, more or less. There's your bandwidth saving.
The conversion to e.g. a NURBS representation is semi-automatable; e.g. here. But potentially tricky.
Doing something in between is perfectly possible using similar techniques. Most modeling software has powerful topological tools, including "simple" things like selecting contiguous edges and loops etc. Fitting a parametric curve of arbitrary order to a loop, say, would also be trivial. Generating a NURBS surface from a grid of parametric profiles is also possible, and actually data optimal.
So, hypothetically, you would select things like creases, seams, panel gaps, wheel arches, wheel rims, exhaust tips etc. etc. (anything that might "silhouette" in-game) to be parameterised, and the rest simply dumbly smoothed in between (luckily cars are pretty simple like that; although interiors are the real pig).
You store the mesh (progressively), plus a list of contiguous edges and loops (at the top of the hierarchy) that are to be subdivided according to specific rules (parametric) and the rest just smoothed as "normal".
"Tessellation" of parametric curves is almost precisely what it was invented for.
Depends on what their current internal modeling scheme is. I ruled out NURBS previously, but now, having seen the latest models, I'm not so sure, despite how unlikely it is.
To get back to the two birds one stone metaphor I used, if the topology is wrong to begin with for the use of AT (which is the most probable explanation) They would still have to start from the beginning of the modelling process. I don't know if that renders everything I said above pointless or not.
What if they designed the topology to be compatible with progressive meshes in the first place? Their modeling is and always has been extremely clean, e.g. here. It's also very well suited to tessellation (subdivision), being grid-like. It should be obvious how careful alignment of edges will make semi-automatic extraction of "features" for parameterisation much easier.
They "re-did" the Evora, which did not have tessellation in GT6 (list here.) That would be a very good place to look very closely, I think.
You raise the point that the tessellation was "software" based on PS3 but that doesn't mean it isn't still some non-standard approach on PS4, despite the tessellation unit (which under AMD's HSA paradigm, well developed on consoles, could be used for other things, in the same way that graphics APIs were used to do GPU compute in the early days by salting / sorting the input data in the right way to get the right output from the fixed API / hardware ops).
The progressive mesh aspects alone must be handled as before (possibly on the CPU), then perhaps the tessellation unit only handles the "dumb" smoothing, and the stream processors look after the parameterised subdivision.
- 903
- Tasmania
- RobUrabus
OK,
I read that post about 4 times.
My head hurts. But I now must research everything you're talking about.
This is the stuff I like, I work with electronics and programming every day and can see why things gets delayed.
Add to that the complexities we're talking about here, multiplied by 150 cars, 30 tracks, and the fact it all moved, I'm surprised that any game gets released.
I read that post about 4 times.
My head hurts. But I now must research everything you're talking about.
This is the stuff I like, I work with electronics and programming every day and can see why things gets delayed.
Add to that the complexities we're talking about here, multiplied by 150 cars, 30 tracks, and the fact it all moved, I'm surprised that any game gets released.
The most important factor no less! Turns out it isn't the most important factor after all!
See where I'm going with that right there?
And you still haven't shown any actual evidence that there would be a discernible lack of consistency between the premium and super premium models. You've given us ifs and maybes, that's it.
