So a couple of people asked me to elaborate on why GTA V is so amazing in a technological sense. So I did. This might be a little unfair, because GTA V is the only urban sandbox game to make the jump to current-gen consoles.
Still, all of this was basically true when comparing (say) GTA IV to Saints Row The Third, so it all evens out.
While running around the city, I ran into this bit of sidewalk, which gives a little clue as to how they cram so much detail onto mundane surfaces without blowing their texture budget:
That little pair of symmetrical cracksHint: The character seems to be looking right at them. in the sidewalk is a pretty interesting clue as to what they’re doing. I knew they couldn’t be doing this through brute force. It would be completely absurd to try and make these gigantic textures for sidewalks and roads. What I’m guessing we’re seeing is some base texture: Grass here, concrete 1 here, concrete 2 here, and so on. Then there’s an additional layer of “decals” to add cracks, spots, depressions, etc. Here the artist used the same decal twice, and flipped one of them?
Also, the sidewalk seam on the right side of the screen (right above the words) looks odd. Maybe we’re supposedly seeing the grout or mortar or whatever that crud is between sidewalk tiles. But my eye is telling me that’s really two decals that are overlapping when they probably shouldn’t be.
Anyway. You could achieve this a lot of different ways. The point is that the sidewalks blend two layers of information together to make a whole that is greater than the sum of their parts. I described a similar concept in my FUEL video, only this time it’s being used for texture data instead of topography.
Footnotes:
[1] Hint: The character seems to be looking right at them.
Diablo III Retrospective
We were so upset by the server problems and real money auction that we overlooked just how terrible everything else is.
Overthinking Zombies
Let's ruin everyone's fun by listing all the ways in which zombies can't work, couldn't happen, and don't make sense.
The Gameplay is the Story
Some advice to game developers on how to stop ruining good stories with bad cutscenes.
Revisiting a Dead Engine
I wanted to take the file format of a late 90s shooter and read it in modern-day Unity. This is the result.
This is Why We Can’t Have Short Criticism
Here's how this site grew from short essays to novel-length quasi-analytical retrospectives.
Hey, that observation about tiles ties neatly into the conclusion you had about AI on the last Diecast. It appears to use relatively low-level techniques (using several layers), but the effect is practically almost as good as something created with bleeding-edge technology.
could the decals even be, dare I say it, procedurally generated?
(with parameters – in this area use more of the ‘unrepaired sidewalk’ decals, in this area use more of the graffiti decals, etc)
That’s what I was thinking. Given the sheer size of the game, it seems impractical for an artist to hand-place every decal, and I don’t think any artist would intentionally place those obviously mirrored decals that are shown in the screenshot – they’d immediately notice it and think “that’s no good”. My money is on some sort of procedural system that was later touched up, and there are parts that were missed here and there.
I don’t doubt Rockstar could afford the artists to actually build everything by hand, of course… but they’re a business and they’re already making a huge game, they don’t want to increase their workload any further if they can avoid it.
I do think Shamus’ suggestion of the use of texture layering for details is probably correct. This is actually pretty old tech (i.e. the Unreal Engine games of the late 90s did this) but surprisingly, it’s not always used that much, and you can still find many games these days that would benefit from that extra level of layering to break up texture tiling (and it’s much cheaper to have one 512×512 texture and one 128×128 texture, than to have one big 2048×2048 texture, but arguably image quality is about the same).
Agreed. My bets are on a lot of procedural tools in use to put most of that content together. Still a lot of human touch-up, but we already knew cooperation between man and machine was better than conflict… right?
Speaking of which, Shamus, how much of your content is procedurally generated? I’d be surprised if you don’t have an “article seed generator” of some sort.
You are right on RockStar being able to afford to let an artist paint the entire street.
Painting tools/brushes/plugins combined with textures can easily “paint” an entire street, then simply cut it into tile pieces to allow streaming/LOD etc.
In Rockstar’s case they probably designed/mapped the town first so they’d know the scale of it all.
A few streets whee probably changed during design, but my guess is that most was fixed in size/length by the time the texture artists got to work and replaced the placeholder stuff.
One issue with “paintig a street” is the size of the resulting tiles, this can be reduced by automatically having a program find matches or close matches possibly with human aid, the result could easily being able to get rid of half the tiles.
Another possibility is that it’s all shader magic (cracks generated using shader scripts and random numbers).
“Painting the world” is what I’d prefer as a artist could really be original then, but the space needed to store the result would be huge, (Rage did some clever things like finding close matches I think to remove “duplicates”).
“John Carmack has revealed that an uncompressed software build of Rage is one terabyte in size”
Probably procedural and using a PRNG of sorts, that way the tiles look the same if you look away and back again.
I wonder how “deep” that goes. If you go the the next street and back are the cracks the same? Is this per game, so if you quit ad restart the game it’s different?
Personally I’d use load time to calculate these things so it remains consistent while you are playing.
Unless they store a PRNG in the savegame.
It’s also possible the PRNG is stored in the gamefiles, I guess you could call it a static PRNG value?
Also that cracked pattern mirroring is just the effect of true random, random does not equal unique.
It’s also possible that this is a artist “mistake” and the same crack layer was used for two different tiles and the world builder/QA did not notice that these two tiles together could end up looking like that.
I am curious though as to how they are doing this. Then again the game (for PC) will be like 65GB right? So using multiple textures/layers and some PRNG numbers the streets may just have unique tile combinations.
Maybe this is another Shamus graphics project in the making? How to make a good looking road and sidewalk procedurally of any length and making it look like nothing is repeating.
I’d bet that they had a program that slapped decals on sidewalks and then they saved that. Maybe each sidewalk tile has a texture and N decal values. Then after they ran the program their artists looked through the results and manually touched it up a bit. They have maybe 20 32×32* decal textures and maybe five positions per tile, and this comes out to taking much less memory than 20 choose 5 (15504 if you don’t allow multiple copies of each) 256×256* tile textures while allowing just as many unique combinations.
What probably happened with that crack is that the system put the same decal at the northeast side of one and northwest side of the other.
* I imagine the actual numbers are higher these days
I’ think it’s done dynamically when loaded by the game, rather than as combining decals and textures as part of the build process.
What I mean is that when they’re designing the area they assign decal and texture values to all the objects, either manually or with an automatic system or both. I expect an automated system set up most of the minor detail work and artists put in special bits and cleaned up most of the egregious issues.
Then when the game runs it puts decals in the places that the stored data says should have decals.
I think this is one of those situations where application of the Cicadia Principle could come in handy: use all those lovely prime numbers in layers to create a huge number of unique variations of decal details with a fraction of the man-hours needed to produce them.
This reminds me very much of borderlands’ gun system, cobbling together details from multiple sets to make an exponential number of resources. I wonder of this could be used to make something like shadow of mordor’s nemesis system have more interesting traits than immunity to attack X or whatever.
I’m an old stage designer/tech, and this really resonates with me. When you’re designing for the stage, you want to create the illusion of “real” objects, but without the expense of building the real thing. The “two foot thick” castle wall can only be three inches of wall with some deeper framing around the gate and the windows. You don’t care what it looks like from backstage – all that matters are the bits the audience can see.
Sandbox games don’t have a backstage. Everything needs to look nice from every angle (and not just every angle you can reach from the ground – it has to look great when I’m flying a helicopter too).
I’m glad we have procedural sidewalk technology now.
This is kind of what I hope becomes possible with voice acting sooner or later (though I imagine the rights negotiations will be hellacious): Randomized and/or targeted changes in pitch and tone to recorded voices that sound natural yet change emphasis, emotion, and other environmental factors.
Take the sentence: “I never said she stole my money.”
You can emphasize each word and completely change the meaning of the sentence. Such a thing in voiced NPCs would help with verisimilitude, not to mention eliminating an angry diatribe directed at you followed by a cheerful “goodbye!”
I’d even settle for it being used only in specific spots, like including the player’s name in dialog.
The issue with player name is that you get names like “Reginald Cuftbert” do you want it to handle that? A Voicesynth might have rules for saying “Josh” correctly but not odd things, and what would be the first name and last name? And in Japan the first/last name is presented in reverse (which trips me up all the time).
The way most games does it “The Warden” “The Herald” is fine, heck in DA2 it was “Hawke” a fixed last name (same with Mass Effect).
I found that the use of a fixed lastname or a fixed nickname works best.
In written texts found you’d see maybe your typed name which is a nice touch.
There have been the odd game that let you choose from a limited set of nicknames (I think it was a arena champion name or something, forgot which game this was though).
A synthetic voice would be best used for random/generic NPCs.
OR take KoTOR for example, the Manaan/Selkath (!?) language (and maybe a few others like the Infinite Empire race’s language) could have been synthetic, the recorded voice for them is so damn repetitive that after the n’th playthrough you are sick of hearing them talk.
If anything it’s for stuff like that where it’ll be used first (to free up more space for textures etc. and to add variance to NPC dialog).
The actual subtitle text could be used as a sort of PRNG so that the same word would generate the same sound each time but still fully automatic.
A designer/writer could then just tweak how the NPC sounds (tonality etc) and just focus on the text/dialog from then on.
Randomly generated NPCs may benefit even more for a voicesynth, here tonality could be random (but using gender/age profiles like male, female, young, adult, old, etc).
But for the player name… probably not going to happen the next 50 years, actual people that know a person has issues saying their names properly, imagine a computer trying to pronounce it properly? I doubt seeing the framerate dive to single digits during cutscenes would be very popular.
Frankly, I feel like the text to speech is good enough these days that I’d be able to put up with the quirks if it meant freeing up writers to write like they did for CRPGs back in the day (being able to write much longer into the process, not having to get approval to bring voice actors back in for rewritten scenes.) Reserve actual voice acting for situations where its especially important (I guess that’s kind of like what you were suggesting.) Current speech to text can pretty quickly synthesize a voice actor’s voice (even if it can’t capture the nuance of their performance.)
Well, there is autotune which is used by many artists (not just for effect but to hide that they are missing their notes), you can sometimes ear a metallic edge to voices when it’s used.
A space sim or similar might get away with it as they ted to use “voice over comms” with filtering/effects and in those cases you may not notice a pitch change etc.
However, for main character you would not want to do it, especially not cutscenes/main dialog.
And coms chatter can easily be done by having the voice actor do 3 different reads of each word/sentence, a normal, a angry, a scared, maybe a happy/victorious, and that would cover it.
There are virtual Japanese (CGI) popstars that uses pitch shifting/bending/tricks (Vocaloid I think the tech is called).
I’m not sure if the processing power vs storage space will tilt towards voice processing, it’s easier to just use more space.
There are voicesynths (used in Windows Text To Speech) for example that does do this.
Heck this all dates back to the early Amiga days (the Amiga had a surprisingly good voicesynth for the times).
Parameters where used to alter how the voice sounded, I can’t recall if many games used it or not though.
A “AI” character could easily get away with this, and it would not seem odd if multiple games had a similar voice in that case either.
But usually a actor is used (Mass Effect’s EDI and Legion) instead.
Having just spent an hour or so earlier today listening to Vocaloids, I think the technology is up to the task. Have a listen.
They apparently can require manual fiddling with settings to get it to sound natural, though, so it might not be able to support generating entirely new lines on the fly. But having someone fiddle with the settings would be cheaper than having a major voice actor rerecord lines. Licensing is an issue, of course.
Ack, no.
Vocaloid technology has come a long way, but their voices still sound way to flattened and squarewave-y for me to enjoy hearing them sing. It’s too filtered and pitch perfect.
I wonder if it would make their timbre more natural if we just added some low amplitude white noise to each output channel, and then gave them a slow, subtle, sinusoidal de-tuning, to imply that they are breathing through meat to produce these sounds.
-or would more human voices just make them more creepy?
It’s in the uncanny valley, in that video in the start you get a very weirdly done “sympathy” word.
Also I doubt a game like Dragon Age or Mass Effect would be enjoyable with voices like that, and this is singing not talking.
Might work for coms/walkies/phone.
And certain aliens or mythical/fantasy creatures might get away with a “odd” voice.
Certain more artsy or experimetal games would probably be able to better take advantage of it.
One game that could have benefited hugely from using a synthetic voice is Tron 2.0
While inside the computer it would have fit very well.
They wouldn’t have to had it live even, just pre-recorded. Storing audio is not that bad these days.
With codecs like Ogg Vorbis can have a rather low bitrate and still sound good, and Ogg Opus is even better at lower bitrates.
The main advantage of synthetic voices is that you don’t need to get the actor back in, the dialog writer can experiment etc.
I once heard (sorry, pun) about the potential for digitizing and reproducing the voices of long-dead actors and other notable persons. Apparently, you’d need a bunch of “core sounds,” like a collection of unique consonant and vowel sounds, certain vocalizations at various pitches and timbres, etc. The tech story made it out to be relatively simple, though I’m sure making it sound natural is quite a way’s off.
For now, I was just hoping they could at least take a line of dialog and alter the mood or just vary how it’s said enough to make me possibly not notice I’m hearing about arrows in the knee or patrolling the mojave in exactly the same way I’ve heard it before.
I’ve seen some games that allowed for various NPCs who used the same voice files to have random minor variations in pitch.
For example two characters running the same basic AI script would have the same voice actor, but one’s voice would be pitched slightly up or down, to make it less jarring.
It’s not a perfect solution, but it does help you get some more life out a limited pool of resources.
I forsee that in the future voice actors may have their “voice print” recorded an d then that is used with a voice engine that allow near real time or real time synthetic voicing in games.
Main characters may be pre-rendered to provbide max quality but supporting characters and generic NPCs may have passable quality real time synthetic voices.
They already digitize face textures of actors.
Like Kevin Spacey in the last Call of Duty, that really looked like him, and was at the quality level of L.A. Noir faces.
A company actually did this for Roger Ebert a little while before he died so that his text-to-speech generator would sound more like his old voice. There should be some audio & video of it in action floating around, as well as a handful of articles about it. It’s not perfect by any stretch, but it’s a big step beyond Computron, Professional Diction Robot.
Remember though, that languages don’t all use the same sets of sounds, and many handle intonation very differently. Until this kind of tech becomes VERY mature, internationalization will be a nightmare.
Hold on. You’re saying they don’t speak English everywhere?
On the flip side, imagine if this could only be done for select languages based on market size. What if Sony or Nintendo came out with some amazing JRPG that utilized this technology and created this immersive experience, but only in Japanese?
I wonder if that would increase enrollment in language classes?
To be fair nowadays games are mostly localized in subtitles only. Still, custom names would be a real mess.
That’s a fair point. I wonder if this sort of technology would be more or less feasible with highly tonal languages like Vietnamese or Korean.
You’d need a notational system that can accurately tell the machIne how to pitch the sounds, but you also have less sounds to worry about, and it’s probably less nuanced (or at least more rigorously formalized) than languages that use tonal bandwidth almost entirely to convey emotional and grammatical data.
Edit: Teh Spellz
For what it’s worth, this is the kind of problem games industry (at it’s current meta) loves to solve. Multiple isolated small jobs that can be solved by throwing more workers at it. It doesn’t require a thematic coherence or holistic QA checks to make sure it fits together. You can outsource mechanics to India, art to China, sound to US without any problems.
As long as the resulting quality is good and the workers doing the work are properly paid then I have no issues with it.
Deus Ex: Human Revolution did suffer due to this, the boss fights was outsourced IIRC and lack of communication or design oversight messed up those fights.
But yeah. A Globally designed game is just another way to design games or software these days.
Movies and TV have done this for a while now, where a B or C unit usually comprised of local crews film stuff not directly related to the main actors, this allows parallel filming.
And in the case of games, the outsourcing/texturing companies have very talented people working for them, sometimes more skilled than European or American artists, they probably got tools and tricks letting them churn out really great textures incredibly fast.
The manufacturing industry are experts on outsourcing, for example stuff gets designed in Europe then produced in China then assembled in the US then shipped globally.
Very odd but very effective is done right.
Take LCD panels for example, there are only like 3 or so in the world making them, but there is a huge amount of different brands.
Same thing with games, it’s probably the same companies in China or Korea or Philipines etc doing the textures for a large bulk of the bigger games in the recent years.
Heck, even The Simpsons have their animation outsourced, which looked odd when they first did it as the art style/QA did not match that well compared to the original Simpson art.
“We’re sorry. Due to circumstances beyond our control, the script for the rest of the episode was lost on its way overseas. Luckily, the episode was finished by the Korean animators. We hope you enjoy the new ending.”
Maybe just a quick holistic QA once_over, though. Otherwise there’s a risk of stuff like the Watch_Dogs 90_degree train_track turns! :D
http://www.independent.co.uk/life-style/gadgets-and-tech/gaming/watch-dogs-developer-opens-up-about-butchered-game-design-9454511.html
What’s really amazing about that is that a city and its infrastructure could be so butchered in a day and age where Google Maps exist.
I may never have been to the third largest city in Britain, for example, but I can “look around” if I was trying to make a decent recreation of it for a video game.
And it’s got adequate bounding boxes in most civilized places. There is little reason some developer couldn’t license, for example, the data for all of Manhattan for an urban combat game. Google got the interior layouts for landmark building already too, and procedurally doing the interiors of boring ones based on GTAV’s tiling tricks should give you a believable if not 100% accurate interior/exterior environment that’s literally a dozen square miles. Mix in GW2-style WvWvW territory control dynamic and you’re set for probably 1200 players in a battle.
Isn’t the logical conclusion this:
The artists actually only determined a small part of what the user gets to see, and the rest is done procedurally: You’ve got procedural sofas, armchairs, furniture, cars, and (of course) sidewalk tiles.
The artist would only have defined the external shape of the sidewalk and what kind of are it’s in (type of matrial, number of cracks, probability of grass growing in them…), and the rest is done automatically.
Same would go for the store in some unimportant gas station store. They would all have a variation on one or two types of shelves, in some procedural arrangement, and in them you’d find [food items] and [car accessories], [drinks] and [newspapers], where the procedural system will simply put whatever assets it has in a random variation of amounts, using a slightly different shelf layout every time.
The whole chaos can be controlled via some random seeds, and the basics for the assets can probably all be kept in memory during the whole game. So if you’re approaching a new area, all that needs to be loaded is a rather high-level description of the environment, and the shapes will be painted in using the procedural assets, or where it matters storywise, I suppose the artists would be able to prescribe otherwise procedural details.
I’ve wondered for some time why no game seems to do this, but there you have it.
It probably wasn’t easy to set up this kind of system (am reminded of one developer interview about “No Man’s Sky”), but now they have it, I bet there’ll be more of it.
… hope it’ll be used on a game that I actually want to play one day…
I feel like the best solution for this sort of fixed-space AAA game is to have someone plug in the layout and then have an automated system make a scene file with all the little minor details. Then the artists go into the scene file and edit it exactly like they would edit one made manually. Granted, I’m not all that good a graphics person, but I think performance-wise it’s better to have everything pre-made and load it into memory rather than generate it on the fly. The advantage of the procedural on-the-fly stuff is storage space; storing everything that No Man’s Sky could generate would be ridiculously impractical. If everyone is supposed to be playing in the same space and it’s of a reasonable size, space is less of an issue.
That’s mostly what I’m talking about, except in many cases it would be enough to just manipulate things that could otherwise be determined by the random seed. As in: “this particular sofa shall be red and have a backrest of this given size, and stand in this particular part of the room”, rather than going in and modifying vertices and mapping texures. This means you’d still use the same parametric models that everything else is made for, except you mold the parameters to your will.
Ideally, these things should be able to co-exist with completely handmade, non-parametric elements: it should be possible to draw an arbitrary, non-parametric line and have a procedural road follow that, then place a custom pothole somewhere on that road and have the road texture adapt to that, too… They may not have gone as deep into it as that, but now I want to work with that sort of system :)
Given a largish texture the problem is to produce an even larger texture that looks sorta similar. Graph-cut texture synthesis is currently the “best” method for this kind of thing, though since there isn’t really an objective measurement “best” is completely arbitrary.
As a side project I am building something similar where instead of remembering the whole texture a neural net is used.
Sidewalks?! I don’t know how you can make a big deal about the variety of sidewalk and road textures when they put such a lame effort into variety of hookers. There’s only, like, a dozen or so different types of prostitute and only 4 or 5 different animations when you …uh…procure their services? The points they get for making highly detailed scaffolding and well manicured lawns is made moot by my inability to hire a transgendered midget hooker in a catwoman costume. A few different animations of slapping them around afterwards to get my money back wouldn’t hurt, either.
I mean, seriously Rockstar, where are your priorities?
*nods head sagely*
Not bad.
By doing two layers its actually harder than making a separate texture, i know this sounds odd but look at Watchdog they made the city of chicago from scratch, if they did it, rockstar can do it as well.
Harder in what sense? It’d probably eat more GPU cycles and maybe some more CPU cycles, but replicate the same level of detail with much less texture data. If the toolset is any good at all it shouldn’t matter much to the level designers; they click a button to add a feature and then the system either creates a new texture or adds a note to apply a decal. If I were designing the toolset for a layered system I’d probably configure it so the artists could save texture+decal combinations and treat them like one texture while the system tracks the decals.
Now, it does seem to be procedurally generated, but there is no reason whatsoever it can’t be done by hand. For instance, Source lets you manually put in bullet hole decals. The main reason I think it’s procedural is that the instance Shamus pointed out is so obviously unnatural-looking it seems unlikely an artist did it by hand. If you look closely you can see that there’s even identical divot patterns inside the cracks, and it’s clearly not a reused sidewalk tile because the other portions are different. It seems quite likely it was autogenerated and passed playtesting because no one looked at it for more than a second or two.
Multi-texturing is cheap on any gpu made this century, and you can save on binds by using array textures. You might pay a bit of a bandwidth if you want extra texture coordinates per vertex but that’s about it.
On the other hand, blending steep parallax mapped textures is still probably prohibitively expensive; although that’s less of a problem in a third-person game.
This may just be nostalgia speaking but Mafia did a pretty good town.
Mafia: The City of Lost Haven (warning, this is the game edited together as a movie)
Watch that for a bit and then devide if the game is wort playing, but anyway, the city felt good, you cold ride the train, take the gondolier.
If GTA: San Andreas ever had a direct competitor it would be Mafia (rather than Driver).
Mafia also did better as a mafia game than the Godfather I and Godfather II games (though the Godfather II game wasn’t half bad and I’d recommend that one).
Mafia II is I guess a competitor for GTA IV, though Mafia II struggles a little on the story especially towards the end (supposedly there were to be 4 endings instead of 1), but it has the feel of a Godfather III game that was never made.
There are rumors of a Mafia III game and I truly hope that turns out to be true.
If Rockstar was clever they’d buy the studio/get the devs that did Mafia I and II and the rights and use the Mafia franchise to do darker storylines, The modern GTA games work best when they are like GTA: San Andreas and GTA: Vice City, not too serious, a tad juvenile, poking fun at the current times and past times.
The Driver games ended up going the way of Saints Row (weird and crazy), which is interesting as I think Driver started out more serious/darker at first?
But anyway, the Mafia game is very nostalgic and the closest competitor to the modern GTA games IMO (by “modern” I mean GTA III and later).
Here’s a nice video listing possibly the ten best Mafia/Gangster games prior to personally I feel San Andreas, GTA IV and it’s two expansions and GTA V should be in that list too, https://www.youtube.com/watch?v=F9gW9vdxols
Interestingly enough these are all sandbox games as well.
Just learned that Shamus is old slang for Private investigator, thank’s for the knowledge videogames.
Now should I call Shamus for Sherlock now, or P.I. Young or Detective Young or Pee Eye Young?
*ducks behind a chest high wall*
How long will it be, do you think, before Rockstar or even an indie coding company just licenses “Los Angelino,” “New Amster-York,” and “Saint Lewis” from their library of mapped-out polygonal locations the same way actual cities tax/license film crews to shoot movies in their locales?
It seems like it’d save a lot of time and be a neat way to have a business focused on rendering (or at least mapping, the textures could cost extra as an option if the licensee was going to re-skin the place) cityscapes and so on, freeing up dev resources for game mechanics and story and so on.
It’s an interesting idea. If the same city could appear in multiple games, you’d have people that could get to know and talk about the city like a real one. The annoying intersections, the headache overpasses, hilarious jumps, etc.
I would actually love to see Rockstar keep the layout of one of their cities across multiple graphics generations.
Not sure but didn’t GTA: Liberty City Stories and GTA: Vice City Stories kind of do that? Reusing the city from GTA III (Liberty City) and Vice City.
And while GTA IV: The Lost and Damned and GTA IV: The Ballad of Gay Tony was both expansions directly using the GTA IV Liberty City map. (there where a few enhancements to the engine) And was also available as a standalone (with both DLCs) in GTA: Episodes from Liberty City.
I have no idea how much of the old GTA III Liberty City map they reused if any.
But my guess for the future is that if Liberty City is revisited in say GTA VII then the map is probably based on GTA IV; GTA is so huge now that it would be wasteful not to use the old base assets, textures, colors and “construction” changes would feel natural (be it for realism or gameplay reasons), so the city would evolve.
(for those curious I’m guessing that GTA VI will be based in Vice City and GTA VII in Liberty City, changing between these three cities/areas kind of makes sense IMO)
Shamus, how did you look at this image so long to notice the mirrored texture on the tiles of the sidewalk… and didn’t notice the repeated textures for the street cracks? That’s the first thing that drew my eyes in when you asked us to look at the image. I was sure you would also mention it.
I wonder what’s going on there. There’s certainly a large texture, but one of them is used on a different angle than the other, and yet I can’t see the seams between the two textures. Maybe if the texture repeated a third time, but the screenshot ends precisely on the spot the texture would repeat this third time. Maybe they have this large texture and some kind of logic program that connects them in a seamless way? Or maybe it is just that the artist cleans up the seam.
Sure, I’m most likely just talking non-sense, since my knowledge of graphics and programming is limited to what little I can remember from your coding articles.
EDIT: Now that I looked again, the street itself seems to be composed of a self-replicating texture. It is only the case that the texture is so large and detailed that it is hard to notice, but I definitely see the same white spots and dark spots repeated three times in the screenshots. Interestingly, it is a different texture than the ones that compose the crack’s on the street where it meets the sidewalk, and at a different angle, but both seem to have the same length.
Yeah a large one then multiple smaller ones sort of layered or alpha blended, giving the illusion that it’s more detailed tan one might think. The GTA devs knows how to squeeze a lot of stuff into one tiny area.
Then again, they had GTA IV and GTA IV: The Lost and The Damned and GTA IV: The Ballad of Gay Tony to hone their skills and really get to know the GTA IV engine and come up with clever texture tricks/methods.
It’s probably that each streets has a huge texture dedicate to it, giving each street an identity.
BTW the guys behind FUEL made The Crew.