Content creation requirements are going insane

Normal mapping is definately in demand at the moment, and to me - once you get down to the nitty gritty of creating the normal maps, theres nothing really all that different from building and texturing a bog standard model. It does take a wee bit longer, but as would any other additional process on-top of the usual Model + Unwrap + Diffuse etc.

Normal mapping isn't terribly difficult to work with, and now there are many different avenues one can take when creating such content (using a high and low res model, generating a texture from a model, "hand painting" details etc).

I think the problem is tied with the other thread concerning photorealistic rendering: the textures have to look more and more realistic (ignoring items such as lighting models etc, as they're not really "content").

I was thinking about this the other day, after watching numerous E3 videos, and comparing them to one another. What are some of the things that make some of the upcoming games (Doom3, HL2, Unreal3 etc) look so good? Okay, we've got much higher poly counts, so the meshes can look a lot more like they should. Lighting models are far more advanced, too. But one of the things that struck me most (and only occured when looking at the videos from lower-end companies) was the lack of repitition. A 512x512 texture thrown on a huge quad looks very poor. Sure, this effect is reduced by making the scene dark and throwing in a couple of lights, but I was really awe-struck by the "apartment" scene in the latest E3 HL2 video, which was quite devoid of the all-too-familiar texture tiling.

What am I getting at? Content creation doesn't just require quality, but it also requires quantity (in a good sense). Game worlds are getting larger and larger, but if they all look the same, they tend to lose their appeal. So not only do we need great looking textures, we need lots of them (which leads us into storage and requirements issues, but that's another problem all together). We can use techniques such as multitexturing to get around this problem somewhat (ie. throw a high-res tiled texture on a wall, then randomly place "details" over the wall to change its look etc).

Then there's the issue of creating photorealistic textures (or at least, close to). Sure, we've got a nice host of image-capturing tools these days, but when you're doing techniques like normal mapping etc, it's difficult to capture a texture that is completely devoid of lighting and shading detail.

And last, but by all means not least, is the issue of visual content apparently substituting gameplay content. It's already happening, and with a greater focus on almost-photorealistic graphics, it's all-too-easy for companies to lose sight of the gameplay.

I'm not sure what this rant was getting at, so feel free to pick as many holes in it that you want :)

Just a geeral rant in relation to this sort of stuff...

I agree in that content will be more a case of more time and resources to create things that are up to speed with the power that computers/cponsoles have. a 400 Poly with one diffuse map takes a HELL of a lot less time than a high poly model normal mapped onto a low poly model with several layers of mapping detail such as bump, transperancy, specularity etc...

If you find someone who is a pro at ALL that...then you'll be doing pretty well. Although there will always be a push for those skills in the one person.

The thing is, to do all that takes time. And time is not cheap. So production for any game is going to go through the roof, just to create content, let alone make it GOOD content AND gameplay.

Which comes to my second point. Power does not, and never will, replace good gameplay. The point in regards to repetition is a good one, a repetitive map gets very boring and unsatisfying very quickly, but I find personally the little touches really make a game. Good texturing indeed makes a good game, but all the little tie ins, and side bits that mean very little to the overall game but flesh out the game universe make a WHOLE lot of difference. Even on a subconscious level.

But then again, you need a solid game base to add the extra bits TO...so it's an increasingly tricky thing to succeed in creating good gameplay...especially ORIGINAL good gameplay.

I'm all for refining what we have created...putting the latest tech to some of the oldest games could work quite well in a lot of cases but I still think there needs to be more creativity rather than remakes with the power we have at hand. So a sort of 50-50 distribution of remakes to improve what exists (remakes, sequels) and the other half people trying to produce as original content as possible.

But back to the point, yes...the games industry IS going to be one insane business soon. But it seems there is a lot of people out there wanting to be IN that business...if the industry can afford them all is a different question :)

I have gatherred there is a growing use of contracting for smaller developers so that people are only employed when they are needed. This may be one way to go for Australian content creators with a growing Freelance content creation industry seeing as the smaller companies need to grow at their own pace in order to make a break into the 'big time'...but we will see.

If developers can hand out their work in partitions, that content MAY be more solid in it's quality as there will be more specialised skills towards that partition. eg. dedicated character animation groups etc.

I'll err...shutup now :)

that is a good point discmage>> i think you are right... there will be a lot more freelance content creation happening.......( if can be afforded ) ........ i have never been employed with a studio thus far.. but am definately coming up to scratch with my skills i feel,,, allready i toy with the idea of working from home as content creator... liasing with a senior designer from some studio perhaps and working my butt off finishing a series of trees eg for a forest scene.... and never stepping foot inside a studio.......
i think the work will have to be sourced from alot of different places to bring together high end visual content for new games..... interesting anyways.

i wonder if this high demand on art content may mean that artist could actually license their ready made content out to several companies??

does anyone know if this is allready happening?? it makes sense to me.

this sort of thing may be needed to allow companies to focus more on the playability of thier product.... and not focusing 6 years on pumping out semi-cinematic visuals for a game that has all the game play inovation of lawnbowls....

While reading this discussion two thoughts had popped into my head, the first was:

We should try and get away from using texture maps on object - by texture maps I mean for colour, since from teh start they were made to simulate the details of a 3d object by using colour. But now we have normal, bump, displacement etc maps that we can use to create details, so we don't need the colour ones now. Plus with all the lighting it might even be beneficial to remove them. So therefore objects/polygons will be singly coloured, then bumped, displaced, and lit and then they'll look good. However I don't really see this working for static geometry, only really for characters / objects.

The other thought that I had running through my head was using all procedual texturing for static objects, since these are things like walls, floor, ceiling etc, it would be better to generate the texture procedually than having to blend several textures to create a similar effect. Plus with procedual textures there shouldn't be any repetition, and as a bonus you should be able to "carve" / damage the static objects and the texture could be easily recomputed.

I mean if we implement these thoughts then the artists wouild be required for creating a "script" (with interactive graphical tools) for the level, and then only be required to build the complex textures for models - barring the colour texture. How does that sound then?

As much as procedural textures would be good on the look and randomness of things..I wouldn't expect designers to script them. I know very few...actually no, true artists wha want to code as well. So you'd be placing a design/visual thing in the hands of a coder...which is fine, as those two go hand in hand nowadays (look at the year of coding involved for the whale's interior mouth scene in Pixar)...but still, it's a thought.

But why haven't they implemented it befreo? It's not THAT intensive on CPU/GPU is it?

The other point in reducing the amount of texture work down to what it actually is...sounds good, but the methods used now and for the next gen (after perhaps might suit what you have in mind) do revolve around a mixture of texturing and such.

The thing is Normal, bump, and diffuse are all texturing skills anyway. Even if you take the need to add your own highlights to the diffuse map as such, you still need to make the bump map to make the light effect it properly...and the same with normal map, as in essence that's just an extension of a bump map (just a damn cool extension :) )

And then you have the very base colour of the texture...and that in itself takes time to look good, to add all your cloth etc.

One method that DOES come to mind however, would be a way to make the light effect materials/objects in certain ways. Much like setting radiosity amounts in 3ds MAX. But in essence you'd be making a specularity map (or equivalent) so it all really comes down to texturing anyway.

Unless, of course, I read you wrong :)

Persoanlly I don't see any other method that is CPU/GPU friendly that gives designers as much creative freedom as they have it now. For me, Unreal Engine 3.0 pretty much has it all...I personally can't see a much better way to do it. without restricting creativity.

normal maps are great but i dont think you need to go to the extreme lengths that those big companies do. sometimes normal mapping can look kinda crappy too, they always seem to have really thick blacks and over glossy highlights, such that in some scenes i just find it awful to look at.

the worst part i guess is how it is likely to blow out production times. yeah normal mapping everything might look great, but do i want it to look great in five years time or do i just want to play it now??

It doesnt take all that long to create a normal mapped character, I'm quite sure alot of people dont realise whats involved.

Creating the normal map isnt time consuming at all, its generated by one of several programs and methods with the input of a high and low poly model, and in some cases you can extract the normal map data from ONLY a diffuse texture.

To give some idea on time using zbrush2.0 + maya our high poly artist can produce a quality high poly model for normal mapping purposes in just a few days. Sure spreading that out for 100+ characters in a half-life 2 type game would add up, but the point im trying to make is its not JUST normal mapping that takes so long, theres numerous other texture layers that take just as long again.

Its a funny thing, the engines and cores behind alot of these games are always produced first, the art, quite frequently comes last, and a fuss about a game taking so long to come out is created and generally aimed at artists, producing art for the engines that are capable of handling massive amounts of data.

Multiple 2048 x 2048 textures on a single player character ? Good god! a single 2048 x 2048 texture is enough to texture 64 characters on the average ps2 game!

LiveWire - a normal map thats black and and contains over glossy highlights is a pretty poor combination of normal + specular mapping, probably the result of time vs quality.

The old 2 out of 3 - COST / QUALITY / TIME - mr money bags can only ever have 2 of those 3, Low cost , High quality, and super quick turnaround, will never happen.

Neverwinter Nights a 4 1/2 year dev time, came out tops. Blew the budget, but everyone got a fantastic game. Lionhead / Big Blue Box with fable 5 years dev time, I can only hope it does the same. Most console / pc games that arent sequels or follow-ons from previous titles that are produced in less than 18 months end up in the bargain bin upon release.

Is there anything to learn from these kinds of experiences ?

Its a tricky art to balance time available vs cash available vs quality. The game designers out there should know this, game design documents are always being changed & modified, almost right throughout the development process, a meeting is had, someone suggests a cool idea and BAM a possible month is added to the dev time to cram it in.

I say hooray for 8 year development times, maybe the $$$ people are finally realising how much time a good project takes and costs. On the flip side, i say a double-BOOOOO! for a project that requires 8 years and ends up pure crap.

/end rant

Discmage: I said that the artists would have tools to help them make the procedually scripted textures, sort of like what ATI's RenderMonkey does to some degree, but taken further... So the artists would be using sliders, buttons, values etc to compose the texture, and then be able to view many large sections of it etc.

And as a solid rule for game programmers - (I think Laura Fryer mentioned something like this) - make good tools for your artists and designers and then the games will flow. So basically it is the programmers duty to create a good canvas and brush and let the artist roam free with it.

hear hear!
or here here ........ whichever 'hear' it is

[quote="HazarD"]LiveWire - a normal map thats black and and contains over glossy highlights is a pretty poor combination of normal + specular mapping, probably the result of time vs quality[/quote]

i agree, i've seen some realy nice ones too. funnily enough those some of the ones i dont like have been in Doom 3 and the UE3 demo!

i would have thought it would blow out production by a long way, seeing as if you wanted to normal map everything you would basically have to do every model twice.

Livewire : Have to agree there, doom3 + unreal3 have nice looking textures, but i urge people to check out Chronicles of Riddick, for some SUPERB textures + lighting. Possibly the best in my oppinion, and probably for some time to come i believe. Doom3 and Unreal just dont exist yet and therefor dont count [:P]

Theres a few different ways of normal mapping, a few of the programs like nvidias Melody, offer a multi-res type scenario using various stripping techniques to produce a low res version based on your detailed high poly model, ( which I might add dont come out too badly at all ) so all you need to do is basically tweak away, rather than build a low + high poly. Still traditionally, if you want a nice clean mesh that deforms well + a nice crisp normal map - youll need an artist working on the highs [:P]

There's quite a few different methods of building normal maps out at the moment. There are the high/low poly sort, which build a normal and displacement map (e.g. SOCLAB's ORB). There are a similar sort, except they take a displacement map as well as the two models, and produce a normal map (e.g. ATI's NormalMapper). There is one which takes a (basic) 3D model, and generates a diffuse and normal map (it's intended for generating wall textures. Called WallBump).

I'm sure there are (or at least, will be) tools for aiding the "hand-painting" of normal maps (by means of editing a displacement map, etc).

Kinda straying off topic now...