,
33 tweets,
14 min read
Read on Twitter
Wanted to do a little thread this morning on the history of 3D video game texturing. We have come a long way since the early days of real time 3D used in home consoles but there are still practices used today to make game textures that go all the way back to those early days. 
Let's get a few basics in order first. Real time rendering vs pre-rendered. Real time is what most 3D games use. That where your machine's hardware is drawing the image in real time. Pre-rendered is using a lot of computing power and time to make one frame.
Because of this you'll get different levels of quality. Games need real time because of interactivity. Static elements like cinematics, or static backgrounds could be pre-rendered. The differences used to be staggering. Heres a prerendered BG and a real time character circa 1999.
Pre-rendered allows for a ton of expensive rendering features that can take hours or days to render on a single frame. And that's fine for an image or a film. But games need to render between 30-60 frames a sec all the time. So you would see a lot of shortcuts in early 3d games
For the 16 bit consoles, you had Star Fox as an early example of real time 3d but you also had Donkey Kong Country that utilized prerendered CG converted to sprites (with heavily simplified color palettes). Nothing would look like this in real time for a long time
Once we get to proper 3Dconsoles like N64 and PS1 you see what real time was unable to do. You couldnt use lights to bake shadows or lights into a scene, no material response, no "bump mapping", and low resolution geometry and textures. So how did artists get around this?
For one lighting information was either painted into the textures (shadows, highlights, depth), painted on each vertice of a triangle, or a combination of both. Player shadows usually were a simple texture that followed the character. You had no ability to cast proper shadows
You could get some really basic shading on models but that was usually devoid of proper lighting info. Stuff like OoT or Crash utilized a lot of lighting info in their textures and vertex painting on geometry to make areas look lighter, dark or tinted a certain color.
Theres a lot of great creative work done in this time to overcome these limitations. Painting or putting lighting info into textures is still done today to varying levels. But as real time gets better the less needed a technique like that is required.
So as you can see theres a lot to build off of for the next gen hardware. The next console generation of the PS2, Xbox and Gamecube would try to help address some of these. First to get a noticeable bump in quality was texture resolution and lighting.
Silent Hill 2 was one of those defining games. The biggest takeaway from this 2001 game was its use of real time shadow casting. This meant some of the lighting info put into textures could be removed but for most of this generation they were still heavily used.
The resolution was what made the game and others of the era. Because of higher pixel count you could store a lot more micro detail in them. But, this was all just in the color/diffuse. Bump and specular mapping was rarely scene. You didnt get proper material responses.
This was another reason why baking info into textures was still common. In prerendered this wasnt an issue. A shitt looked like cloth. Glass, hair and skin could look believable. Real time needed a bump. And they got some BUMP near the end of the generation (only on xbox)
Specular maps and normal maps came onto the scene in games like Halo 2 and Doom 3. Spec maps let surfaces react to light closer to how they should. Metal was shiny, etc. Normals let you put more detail you could never get away with putting in such low poly count objects.
If you work in 3d now you know what a normal map is but for those that dont it's a type of bump mapping that lets surfaces react to light with more detail that what is actually in the model. It's a cornerstone texture used on pretty much every model in a game following this gen
With the introduction of this, you saw a lot of game artists change how they made textures. With normal maps you needed to spend significantly more time making an asset. Sculpting tools like Zbrush became the norm to make high poly models baked to a texture to use on low poly.
Before this a lot of textures were either handpainted, bashed together photos in photoshop or a combination of both. With the 360 and PS3 era, that way of making textures went away for a lot of games, as the quality of assets went way up with the resolution bump
We also got a huge leg up in how materials worked, with pre-compute shading. This was a game changer for a lot of artists. Materials could be far more complex than before. This demo in 2005 was unlike anything seen at the time. The 360 wasnt even out yet.
We also got the introduction to a new way of helping to light a scene. Ambient occlusion. Real time had to play catchup again. It was too expensive to render real time so artists just put it into the textures! AO recreates indirect shadows from lights, something too fine a detail
Even today real time AO isnt 100% there for real time but its close!. It's gotten better with processes like SSAO or DFAO compared to even 10 years ago. Baked AO maps are still used today but will probably go away at some point as renderers get better
To sum it a up, with the PS3 and X360 era you saw much higher resolution bumps once again from the previous gen, new textures introduced for shading surfaces. And of course, lighting bumps! You could get real time shadows for a whole scene or bake the lighting for more detail
All pretty great right? Well there were still some drawbacks. Low resolutions on models and textures, plus a lot of overhead for the new shaders AND dont forget resolution in terms of what the games were outputting. 720p! (And tint font on a CRT TV became an issue)
The other issue we had was specular maps. During this time you had one map for how "shiny" object was. This was a limitation ultimately. Materials across the board didnt feel real. So you started to see some people separate out the specular maps. Bioshock Infinite being one
Specular maps would now be separated between one for what material type it was (wood, gold, concrete, etc) and one for the "age" of the material (scratches, wear, etc). This coincided with a new type of shading model, Physically Based Rendering. PBR.
This brought us up to today and the current generation. PBR became standard for a lot of games. A technique popularized by Pixar, it standardized a way for making believable materials in CG. And it could be applied to real time!
On top of that, the industry built off of a pipeline that came into play in the previous generation, screen effects. Things like tonemapping and color grading got enhanced in the current generation. Something that in previous generation you would spend ages adjusting textures for
Theres a lot more going on in this gen but I have to get going now! I'll try to pick up later but hopefully this was something you found interesting!
If you're interested in learning more about older game set and rendering techniques I highly recommend @dark1x's DF Retro series on the @digitalfoundry channel. He does fantastic work looking at individual games like Silent Hill 2.
Just to close this out I'll show you what textures it would take to make an early 3D game versus what what it takes to make one texture set in a game today.
Since this is kind of blowing up, I'll add this little bit. Games are made by people! The tech and processes are always changing but we should still appreciate the accomplishment's artists achieved in their era and how it may have helped build what we do today. Appreciate history
Hey! I'm back! Let's talk a bit more about how some of the techniques developed for older eras are still applied today! Theres a lot of "stylized" texture work that use lighting information in textures. Most notably Blizzard uses this
They blend both a technical constraint with strong art direction to get some fantastic results. These may not follow the same mass amount of textures to get the compiled result of other AAA games but you cannot argue the results.
And sometimes using the PBR technique with some handpainted/simplistic textures can really go a long way. Having a modern engine with tons of rendering bells and whistles helps too I guess.
