MegaTexture Hype

Is it just me, or any other of you think that the Megatexture technology that the advanced DOOM 3 engine is going to provide is just nonsense? I mean besides the first awe of the hypie name, when you settle to actually think about it the megatexture technology is just a pain in the @#%!

For those that are not familiar with the term “Megatexture”, megatexture is the new killer feature of the DOOM 3 game engine. It is about using a singe huge (where mega comes from) texture for texturing the whole terrain. This is different from what is commonly used today in game engines, where artists use a set of tile-able textures that repeat over the terrain. So with megatexture the artists are provided with a single 32,768 x 32,768 image that is stretched over the whole terrain and start painting it. Check these links for more information:

I must admit that the concept sounds very promising when you first hear about it. It seems like the way to go for future generations. You no longer have to see the same pattern repeating over and over again. But think of the artist! Do you really expect someone to uniquely draw such a huge texture? And if he does where is he going to do so? In photoshop? Well try opening a 32,768 x 32,768 image in photoshop and think again.

There must be a custom tool to handle this kind of textures, made by Id. And I’m pretty sure that this is not going to be so elaborate as photoshop, no matter how hard they try. The only way to have the artist work with good tools is for the custom tool to export small parts of the terrain texture to photoshop, let the artist draw in it and finally reimport that in the custom tool. This not a good artwork pipeline by my book. It even messes with the way artists work beyond export/import.

It is quite logical that the system will provide a procedural system to layout the terrain and then let the artists refine and add detail to that. This can easily be simulated with decals over an existing terrain implementation.

I have seen videos from the upcoming Quake Wars:Enemy territory that is using this technology and as I expected there was nothing impressive about the terrain. It actually looked very low detailed compared to what I have seen until now. Which is quite normal if think about it. A 32,768 x 32,768 texture starts to look pixelated over a big enough terrain and it will never have the quality of a really huge terrain that is rendered with detail textures. You can paint any detail on the ground but it will never have the fidelity of a 512×512 decal texture placed on the spot of the detail. Simple as that…

The conclusion is that megatexture is a technology that burdens artists will a “mega” amount of work, that in the end does not result in visual improvements. It also uses lots of resources and adds more restrictions to what you can do. How big can you go with a 32,768 x 32,768 texture? Can you do World Of Warcraft with that?

Megatexture is just a nice thing to implement for the hack value of it but I don’t think that has to offer anything to a game…


  • xdeon

    you misunderstand the concept of megatextures. it’s not painted pixel by pixel, and it isn’t just for texture maps. the artist sculpts a 3D landscape with 3D painting tools and bakes this to texture. this texture is one of many channels in the megatexture file, some contain material data (for sounds and particles) while others may contain trigger information or a vegetation map of sorts.

    The idea is not “making it all by hand” rather a better way for dynamically storing huge amounts of level information in one file.

  • I think i have a good understanding of megatextures since I have a protype running in my game engine. All the “sculting” and “baking” you mention, are part of any terrain system. Megatexture is not about that. Its about using a single texture. period. It is amasing how people fall for hype! They (Id) talk about “stroring material data for vehicle friction, footstep sounds” and stuff and people go wild like this was never done before and its due to megatexture! I meen come on… all games since the eve of gaming have terrain properties that allow for differect footstep sounds…

  • Random

    Actually the entire reason why megatexture is so hyped is because people write articles about how ‘impossible’ it is to edit and overburdens artists because they think they’re going to have to load a 6 gig texture into photoshop which is so obviously not the case. But if that’s how you implemented it in your game engine then yeah, that’s not good.

    No, it can’t meet the demands of an MMO (which essentially requires an endless amount of terrain to work with) but that isn’t the point of it. If you already know that megatexturing is overhyped then you should be able to see that it is not the end all solution to terrain and should not be viewed as that. It is nothing more than one new method of accomplishing the same thing.

  • Defrag

    Here’s a couple of thoughts:

    1: I wouldn’t be surprised if the megatexture is assembled on the fly as a level is compiled rather than requiring manual manipulation from individuals. The notion of artists manually cutting and pasting into a 32k x 32k pixel file is absurd; in fact, it’s so absurd that I don’t know why you brought it up. At the worst, megatexturing should offer the content creators exactly the same efficiency & workflow as traditional systems. Potentially, it should allow totally unique textures in all areas of a level (whether people have the time or money to do this is debatable).

    2: Carmack’s implementation hasn’t been examined properly yet, so how can you dismiss it as being as unworkable as your own? The chances are that Carmack’s implementation is integrated with the engine and tools and provides a seamless abstraction meaning a lot of what you’ve ranted about is null and void.

    3: I get the feeling that some of what he has talked about regarding “potentially unique everything!” has been overblown and this is where the backlash is stemming from. However, Carmack has apparently developed megatexturing beyond terrains and it now represents a seamless paging system that handles all texturing in the game. To me, this is infinitely more useful than terrain-only mega-texturing. As a level designer, I’ve often had to go through a level removing textures here and there so that my level used significantly less texture memory. I.e. I have removed resources from maps, making the maps look more “samey” and uniform due to memory constraints. If mega-texturing is as versatile as it sounds, those days are over. It should hopefully remove these memory versus uniqueness tradeoffs. That, for me, is the beauty of the system, not the bluster about terrains with hundreds of uniquely painted textures that are all slightly different.

    In short, let’s wait and see. Part of your argument may hold, but I feel it’s totally unfair to criticise the implementation & workflow when you haven’t tested Carmack’s implementation. I’m sure I could implement it too, but I’m not John Carmack so his implementation will likely be several times better than mine.

  • defrag I think that you don’t really get hardware. If you had a map that you had to cut down textures to use less memory, megatexture is not going to be a solution. The map view that was swaping textures in’n’out with your regular texturing will do worst in a megatexture engine. In a megatexture engine its like having unique textures in all polygons, which I’m sure wasn’t true even for you most textured map. Its just that the megatexture engine will cut down the maximum dimension texture you will have in the view.. which is like lowering the texture quality in a regular engine. That is the cause of big texels that can be seen in QWET near the view. That is about the “beauty of the system” you mention. It made you think that you can texture beyong you cards memory capabilities, which is part of the hype and far far from the reality.

    random: no I’m not using photoshop in my implementation, i use a very fine tool (third party) that automaticaly generates the terrain.

  • Drew

    Just because you have a different mockup running in your ‘game engine’ doesn’t mean you’re producing anything like the QW megatexture. And basing everything you know about the technology off of some low-res movies is pretty silly as well.

    Just because you haven’t found an amazing way to make it work like John Carmack and Splash Damage have, doesn’t mean it can work. Is your game engine on par with doom 3? Did you build it from scratch? How us some screens!

  • Sylphis3D Developer network

    What carmack did was not black art… is well a known technique…

    I’m not basing anything on low-res movies. Read the post again. You are not even right about the low-res part… there are high-res high-quality movies publiched by splash damage out there, you must have missed it…

  • JJ

    You, my friend, need to read up on MegaTexture again. You’ve got it wrong on so many counts that I can’t even be bothered to start correcting you. If you pay attention, you’ll see that there is a tool that synthesizes the initial megatexture from a set of source JPGs, using rules to place textures in the appropriate place. This means snow on mountains, grass on fields, mud on river beds etc. The artists can then go in and ‘touch up by hand’ specific individual details using a custom tool.


  • Pingback: Thoughts Serializer()

  • Rimtiprop

    uhmm, what the heck are you guys talking about? Yeah , you dont understand megatexture very well, the point is to REMOVE texture limitations and to reduce draw culls by having all textures on the same sheet. Id tech4 does not have megatexture implemented properly(its more of a customized doom3 engine as you say) , Id tech 5 on the other hand is.

    I suggest you actually TRY building something with the tools first, which in your case wont be until its released. Ohh btw there is no loading between areas with the next gen id tech, that means seamless megatexture areas together, so its perfect for a MMO.