texturing in the early 2000's - how did you do it?
null
this post is directed towards the veterans/old timers that were around in the early 2k/late 90's.
i was wondering how the texturing work was done back then. mostly referring to games like max payne/crazy taxi/blood2/kingpin/bully/manhunt etc.
i was wondering if there was a texture authoring tools like today. or people simply had scans/photos/ where they be mostly hunting for textures.
some people here already mentioned photoshop. but those kind of techniques are lost in time.
if any of you want to elaborate, ill be glad to hear it
i know working in diffuse/specular workflow only can be tough, but its a style that lots of people love and i know some people who will try to replicate it. it could be that most of you are new but maybe there are some old folks around
i was wondering how the texturing work was done back then. mostly referring to games like max payne/crazy taxi/blood2/kingpin/bully/manhunt etc.
i was wondering if there was a texture authoring tools like today. or people simply had scans/photos/ where they be mostly hunting for textures.
some people here already mentioned photoshop. but those kind of techniques are lost in time.
if any of you want to elaborate, ill be glad to hear it
i know working in diffuse/specular workflow only can be tough, but its a style that lots of people love and i know some people who will try to replicate it. it could be that most of you are new but maybe there are some old folks around
Replies
I consider these 'antique' techniques that are 'lost in time' (c'mon... 2-3 years is hardly long ago) to be still the very basic foundation. Not to mention that mobile games are still using a lot of that stuff.
You can probably find pretty much everything about this and more in the polycount wiki.
Serious answer:
Unwrap your UV's manually and render a preview as an image file. Open that up in Photoshop and start painting on top of it. From there it was paint in PS > check your work in the editor > repeat.
Specular maps weren't that big of a deal because you can still paint a spec map by hand, but they were introduced into game engines at about the same time as normal mapping. Normal mapping made things tricky enough to warrant the necessity for dedicated texturing tools.
Regardless, many programs like Maya and Cinema 4D had painting tools built in but Photoshop was faster and more robust, even if you're working on a flat surface. I think Mari's been around for a while too.
for reals though, Im so glad you dont have to paint on the actual fking UV Map any longer. painting directly on the model in a 3D viewport is a godsend.
but yeah basically what everyone in the thread already said. I agree with lotet though, the real god-send was when projection painting was more available. I remember trying to track down Deep Paint 3d after seeing a video of it and was never able to get it working for me, and development had been stopped I think too. Thankfully 3D Coat came out with that shit, easy as butter. There's many options now as far as I'm aware.
the photoshop techniques are definitely not lost in time, however
...but besides my acient background, I still expected this to be taught - as it is quite basic knowledge I would've assumed.
I took a lot of reference photos myself and used them to avoid having too much of my stuff look the same. Another company stole some textures from a game I worked on and I spotted them right away because they were based on photos I took.
For the PS2, you tried to make all of your textures fit into the 8k texture cache. To do that, you made most of your textures 128 x 128 and 4 bit (16 colours). 128 x 128 x 1/2 byte = 8192 bytes. You had to avoid too many colours or your 4 bit palette would be ugly as hell. You also tried to tile your textures as much as possible to create triangle strips but still break it up in a way so that it wasn't obvious.
The proper way to do it would be hand paint everything with a 16 colour palette but when you have five months to make all the environment art in a AAA game you have to cut corners. I worked with some character artists who hand-painted everything and they did a great job at making a single diffuse look beautiful. Often, the game couldn't render their textures at a high enough quality to appreciate them.
Two photo-based textures I made around 2004 for a PS2/XBox game.
Later, when you had to ship on multiple platforms, you couldn't make your textures just for the PS2. You had to make them for XBox too so you tended to make the textures higher resolution and re-size them for PS2.
I would sometimes get fancy and make textures procedurally. I would also create procedural base layers and use photos on top to grunge them up. Below is a procedural texture I made for a PC game started in 2002.
There were tools like Deep Paint but I never saw anyone use anything except Photoshop. There were some custom tools used to reduce images down to 4 bit and re-size them on each platform.
Mobile has some similarities to the old days but is actually much easier. PVRTC compression on the iPhone is 4-bit by default, but it does it totally differently than having a single 4 bit palette for the entire texture. The next generation of phone games will probably be PBR so the diffuse/spec/normal workflow will be totally dead soon. This is a good thing.
When I was a lighter, I would get textures with all kinds of value ranges. One group of people were convinced that if they crunched all of their values below 0.5 it would make their textures look "better". I could never convince them to have a normal value distribution. You would get random opinions based on their experiences on other projects.
The old days actually sucked hard. Things are much better now.
Yeah, much of the tools and hardware back then is no where as advanced as it is today, but the beauty of development back then was there was a lot more room for taking risks.
You could make games that were hand painted and still pass them off as realistic. There was really an infinite amount of ways developers could innovate because everything was still new.
Even the lack of hardware didn't stop developers from inventing new techniques to get the better of them. Like for example, one developer managed to get bump mapping working on the PS2 (Enter the Matrix).
Or even physically based rendering. Check out the game called Wreckless, and the developers were really ahead of the curve.
Back then, most of the public didn't know about this stuff. So when developers dropped these games with new rendering techniques, it was mind blowing.
We not only have amazing tools today, we get them at ridiculously cheap prices. This has helped thousands of indies making their own games which was very rare 10+ years ago. Unity and Unreal are free which is incredible. You pretty much had to pirate everything unless you were a millionaire. Now, I can get Substance, 3D-Coat, Quixel, etc. for dirt cheap.
I've been using Houdini for the past month and I just had to download an apprentice version, no problems at all. Companies did not do that sort of thing 10+ years ago.
Github and similar services for free? Did not exist. Digital distribution? Nope. Slack? Discord? Goes on and on.
It's much, much easier to be small and innovative now thanks to these changes.
I'd just add to that on top of all that, education about making art for games is more accessible than ever. Back then there were no youtube videos, professionals weren't uploading their shit to Gumroad and blowing tutorials out at $5.
I expect 3DMotive and Eat3D brought a lot of people into making game art who otherwise never would have tried. Does anyone how many Damaged Pillars were posted here after Eat3D released that tutorial? A lot for sure.
Now you have companies like Epic and Allegorithmic doing routine training livestreams with Q&A and youtube channels full of free official training content.
People always romanticize the past but it's rarely as good as they think it was.
2. Make a new 512x512 photoshop document
3. http://cgtextures.com
BONUS ROUND
4. I think I took this book out from the uni library alot in 2008.
*sigh*...completely forgot about that, now you made me miss the old days, not cool >,<
I used to just throw buckets of green paint at my 8inch, 4foot deep green Amstrad monitor.
I had to cautiously re-read the OP's post as I thought it was a pisstake at first.
Here's a random prop I did back in the early 00s, nothing stellar but I still have the image to hand:
All the metal parts were painted or generated in photoshop, the grating, the seams, indented part for the keypad. But then for the keys themselves I just grabbed a photo of a keypad and cleaned it up a bit and isolated just the bits i needed and shoved them together. Same for the LCD display, used a photo for that. All the dirt was painted in, and the sellotape, postit and graffiti was painted also. So I'd say 75% painted and 25% photo. Sometimes it was 50/50, sometimes it was all photo or all painted. All photo generally involved a lot of clone stamping and basically being a hacky bastard non stop. oh my, I'd forgotton the _sharpen filter_!
The primary difference is today there are a lot of standardised workflows, because the shaders are so much more complex you have to work a certain way and it means you can't get away with, for example, clone stamping, so easily when you've got 8 channels to worry about and physical rendering.
Ah if only we could cobble together a single diffuse map today and get away with it / feel satisfied. sigh.
Having said that, I do think there was a warning the game industry should have taken serious, that relates to my point about risk.
The PS2 era of making game art represented an equilibrium of tech and creativity. There was enough power to visually represent anything you want, but not at the cost of requiring millions to develop it.
Fast forward today, and game asset creation requires a ton more attention. Instead of putting all your energy into a 256x256 diffuse, you now have to worry about generating a high poly mesh. Then baking it down. On top of authoring many different texture maps, and analyzing them under different lighting conditions.
It made game development much more complex instead of focusing on crafting the overall gameplay.
I think back to some old threads I made, where I wanted to recreate that feeling of PS2 development. For example, my PS1 open world. The tech was very ancient in 2015, but what I enjoyed most was I was able to quickly put together my game idea and have all the assets reflect that. I thought less about the hardware I was targeting, but more about the type of game I wanted to make. 300 triangles felt like enough to model a car, so how about I move onto creating the car lights? And then when the car lights were done, how about creating a texture atlas for all the buildings?
Now imagine doing that again but what with modern hardware in mind. Each prop would probably take a few weeks to a month and the idea I was trying to sell (an open world car combat/racing simulator) would be lost until the very end.
Again, I do see other advantages that modern game development does bring. Unreal Engine being free is a huge boon and it definitely made game development more accessible to masses. But when it comes down to what the current market has now accepted in terms of visual fidelity and how that translates into what type of game art is popular (i.e the studios who can afford to put out games with rich visuals are likely stuck doing very safe and mainstream ideas, as opposed to freely coming up with their own designs or gameplay) we needed to seriously address how could mid-tier development survive alongside them.
either way, i have interest in the subject because the textures back then weren't entirely handpainted but not fully photo-slapped too. which made things intresting
Photographic downsampling algorithms leave a lot to be desired when you shrink things that small. Important fine lines would get lost or smudged. Some changes in shape would appear too extreme.
You needed to be able to give the impression of more detail than was there in the image. This was very much an artistic process.
You'd also wind up giving important areas more texel density. Such as faces. A lot of things were mirrored creatively. Not always down the middle, but what could be mirrored to reuse uv space, was.
The technique was largely "just paint the thing". It was almost all artistry with some photomanipulation techniques in there to get quick detail.
The realtime lighting tech of that era was limited to vertex lit blinn, or lambert models, unlit diffuse only, or diffuse with cubemapping. Shaders were very simple affairs usually, and reserved to the domain of the programmer, as you had to build them using either fixed function pipelines for the 3dFX/Glide chips, or with a very early version of CG for the early nVidia Geforce line. All of which required programming in either an assembly language or something akin to C. Most of the material techniques taken as standard now was limited to movies and offline rendering that would take days to complete.
Some tricks people would use were early "render to texture" functions from Max or Maya, where you could set up lighting and bake that out to the UV's to get some quick initial shading. Like an AO map but for 1000 triangle models.
We had Wacom tablets but they used a serial port (the 9 pin kind).
I found some stuff from 2006 I did for a polycount contest for a Q3 model. This was a "retro" contest at the time even, with specs matching up with 1999-2002ish workflows, but without the 256 color palette restriction. I wasn't the most skilled artist then by far, but I think it was selected as the default skin of the pack, so its at least a passable example of early aughts texturing. I've attached the full res textures here, and a screenshot of the model, rendered in maya I think. I've also attempted to attach a zip file containing the PSD's for the textures so you can see how it was put together. Its basically just "paint it" with some pattern textures making up the base metal for quick texture detail. Hope that's useful to your quest for historic knowledge.
.
Here's a really old piece I did 'back in the 2000's'
Maybe in 4-6 years, but not yet...