So there was a
thread on Neogaf about about turning modern day video games into old PS1 ones. I posted some old work where I wanted to make an Infamous Second Son scene with the limitations of PS1.
I was told for example, to represent a lamp bulb it would just be two intersecting planes with sprite based bulbs.
I'm
aware of the low poly thread and definitely look at it for inspiration but I want to hear more technical stuff. Like, what are the every day 3D modelling stuff we take for granted that would have been a pain to recreate on 33mhz systems?
Also, it doesn't have to be just modeling. What about texturing and lighting? There was no hardware driven soft shadows, specular reflections etc so how were these accounted for?
Replies
Texturewise, you'd want to index as much as possible. Use 16 colour textures wherever possible (You'd be surprised how much you can do with such a limited palette) and 256 where you absolutely can't. Keep alphas to 1 bit unless you really need a smooth-ish transition. Even then you'd want to reduce the number of colours in the gradient.
Make characters segmented if you can get away with it (you might not have had a choice about that one back on PS1). Fake geometry with alphas where you can - things like car tyres can be dramatically reduced by making the outer flat surfaces alphaed planes and then cutting the cylindrical part right down.
Back then we did not focus on quads for one thing.. Not on the playstation atleast.
So the arch-way would be much cruder, and the lamp posts would be 3 sided cylenders or a sprite on a plane with a camera facing rotation thing so it always looks like it is facing you.
The top piece of it would also be one flat quad for the structure on top of the post to hold up the lights using no camera facing contraint and if you want to go overboard, you put an extra quad at the bottom of it with say a 3x5 pixel stretched texture so when you are close to it looking up, it would not look super thin.
Each bulp on top would then be a quad on its own, with the same thing going on, but technically these could be camera facing on all axis´s, so even by looking at it from above or below, the bulbs would always look like they had volume.
You have to remember that the PS1 had a grand total of 2 megabytes of internal memory to do EVERYTHING.
1 MB of which was usually relegated to a framebuffer.
So in reality, 1 MB of memory for graphics, sound, all the code, handling the loading from the CD rom and such.
It´s insane to think of what we managed to get out of that machine at the time during the end of its lifecycle.
I think Gran Turismo 1 and 2 are still my two favorite PS1 games that amaze me the most graphics wise.
One example of a game for that time that I worked on and remember roughly was Fifa 98 on the Saturn (not far from the PS1 in power but used quads only as it did not really do "3D" but displaced sprites to look like 3d!
The entire roster of players in that game was made of ONE character model, and I splurged a while 100 polys on it.
ALL the team outfits were one 256x256 texture (16 colors), that used color cycling to make every team´s outfit.
The Statiums usually came in at 1000 poly´s total, with two 256x256 textures (16 colors), one shared for all the advertizing boards, and one dedicated for the stadium.
What I mean is, I know how in Unreal Engine 3, you had a hard 60k limit for any mesh you can import whereas in Unreal Engine 4, it's now unlimited (till your PC crawls).
So could the PS1 or game engines back then even render a sphere or was that considered impossible and thus you had to rely on boxes or flat planes facing the camera?
You used as little as possible to get away with the illusion of a shape and not a single triangle more.
A 3 sided cylinder from a distance at 320x240 screen res, looks like its a cylinder and not a box or a plane if its positioned correctly.
4 sided, or even 5-6 sided if it was a big object like say a column in a dungeon in Tomb Raider.
but a small thin object, a 3 sided one got the job done.
This is one good example of poly counts on the PS1.
Tomb Raider was considered a super nice looking game when it came out, with a very nice main character model.
They managed this because the game rarely had more than a dungeon environment, and Lara + up to 4-5 monsters on screen at once.
(left side is obviously ps1)
But with the advent of HDTV and monitors, it's kinda harder to sell that illusion.
Especially for a demake.
I have nothing to add except to say that I heard this entire post in my head in the voice of a grizzled old man with a peg leg and an eyepatch poking at somebody with a cane from a chair next to a fireplace.
And it was awesome.
Gotdamn. I'd love a visual breakdown of this stuff
cptSwing: I do so wish I had any sort of backup of some of the work I did back in the day, but EA was VERY strict about their stuff, even when it was from out-sourcing companies handling platform conversions.
The whole color cycling idea however is very simple, and was used alot on the older consoles and machines like the Amiga and Atari ST.
Basically you would set up something like a few colors (often 2 or 6) and then simply hue change to get different color gradents.
This was also used to great effect to "animate" things like water :P
You could butterfly the road texture too to save yourself some more space and improve texel res too. I'd also dramatically reduce the amount of detail on it to make it less obvious that it's stretched.
Take a look at SkankerZero's entry from the Mutant League contest. He gives a pretty detailed breakdown of exactly how he achieved a similar sort of technique to what's been described here.
http://www.polycount.com/forum/showthread.php?t=91509