Reflections on Flat-Shading vs. Texture-Mappig
or I Can Live with Flat-Shaded Polygons
Just before Christmas, I rented two Sega Saturn games, Virtua Racing and Sega Rally Championship, in order to compare the two. Santa was only going to bring me one, so I had to choose: VR was my favorite in the arcades for a very long time, and I had heard just wonderful things about Rally. So I sat down and played both for few hours each, on separate occasions. I fell in love with Rally, but did not much care for the translation of VR -- it was missing the arcade feel. (though, since Santa also brought me an Arcade Racer steering wheel accessory, I happen to really like VR -- perhaps as much as Rally. The steering wheel makes all the difference in the world with this game... I'll be buying it when I recover from Xmas.)
Anyway, a strange thought crossed my mind after playing VR for a while: Why is texture mapping such the rage these days?
I was totally caught up in tex-map fever myself. It seems that everyone else is too, including the press and a lot of consumers, like the guy in rec.video.games.sega that recommended getting Rally over VR because it was newer technology (read tex-maps), even though they were both released within months of each other. Lately, most games have been judged by many -- most Saturn owners excluded -- on the quality of texture mapping and special effects only, with game play a secondary consideration. My personal opinion is that the original Saturn Virtua Fighter has better game play than Toshenden on the Playstation and was the more enjoyable game for me. I love TSD's graphics, but I'm a long time VF arcade fan, so I bought the game I wanted to play. But oh MAN, did I want VF Remix! Bring on that texture-mapping!
Then it hit me. VR really doesn't look half bad (even though at the time, I thought the control sucked using the standard controller, and had not yet tried it with the AR.) I even kind of liked the clean flat shaded look in this game. It really shows that the Saturn can render over twice as many flat-shaded polygons per second than it can texture-mapped ones. You can see quite far down the track, and get a really broad view in certain VR-modes. The environment is well populated with 3D scenery, like trees, statues, etc. And the frame rate is really good. Some of the objects do things when you crash into them, like noses breaking off of statues, coconuts falling out of palm trees, a snow man that can be punted down the track, and signs that can be knocked over (um... yah, I'm all over the track and crash a lot, that's how I know about these things).
As a professional software engineer dabbling with 3D graphics in my spare time, I realize that a lot of horse power is freed up when displaying only flat shaded polygons and that texture and bit maps consume a lot of memory. That extra CPU time and RAM could be used to make the game much more interesting and animated, as well as use more polygons for greater detail, or provide a wider world view. Oh, sure, you need fewer polys to make an object look good with texture mapping. But take the palm trees in VR. They look really good from every angle. The animated pit-crew is pretty neat. In Rally the trees and people are just card-board cut-outs that always show the same face to the camera. The only animation I've seen in Rally is two zebras that wag their tails and lift their heads, an elephant that walks, and a flock of birds. Fewer polys sometimes means less realism if it means motion is ignored. And it seems that everyone is concentrating on improving visual representation to improve realism, but only paying lip-service to animation as another very good technique. Instead of using (precious) memory to store tex-maps, realism can be added by including animated and intelligent behavior. Time spent improving the visual realism of objects could instead be used to improve motion-realism.
I could like a game that uses very few or no texture mapped polygons if the freed up resources were used to:
- Display more polygons to provide a wider or more distant view
- Include more and more solid-looking 3D objects
- Animate the objects to convey realism
- Add interesting behavior and intelligence to the objects
- Add more interaction
- Better simulate physics
I don't think Rally is a good candidate for this, because it would be hard to simulate what Rally does with only flat shaded polygons: mud, sand, paved roads, etc. would be difficult to represent and are a very central theme in the game. However, VR could, perhaps, have been done even better assuming the Saturn's second CPU is underutilized (and more time was available to bring it to market). For example, the palm trees could have swayed in the wind and branches blown off and down the track, weather added to vary track conditions, animated bystanders and animals could be done, wind blowing bits of paper down the track. All of these would add realism using motion vs. textures.
Where this could really shine though is in a game designed from the ground up to take advantage of the points I mentioned above. Where the world is artistically built to leverage a flat shaded look, and animated objects combined with detailed physics are central to the game theme and used extensively to add realism.
If the same amount of time (OK, perhaps more time) were spent on modeling object behavior than is normally spent on visual representation, the objects could come to life more thoroughly and believably. There could be many more things to interact with, that could be picked up and moved, made to do something interesting, or that act on their own accord.
So, I want to tell you developers, and especially you producers, that I could like a game that relies heavily on simple flat-shaded polygons and only slightly, if at all, on tex-mapping -- if the points above are taken into consideration and assuming I like the game itself. If marketed effectively, with due attention put on the fact that the game stresses motion realism over visual realism, it could, perhaps, even successfully overcome this fixed notion that texture mapping is the only way to go. I, for one, would certainly give it a fair chance.
IDEA FOR A NEW GAME GENRE
This brings me to an idea for a class of game that melds several genres. This class combines RPGs, fighting games, driving games, etc. into a single seamless game. Instead of striving to be the best at any one genera, the game employs diversity and dynamic behavior to make the game interesting and enjoyable. The memory and CPU time freed up by not employing texture-mapping are used to increase the complexity, size, and freedoms of the playing field, and the dynamic behavior of objects that populate it.
As an example, in one of the Bruce Lee movies, the hero had to fight his way up seven levels in a house, confronting more dangerous adversaries at each level. This could be done rather easily with today's fighting-game paradigm. But that is only the climax of the movie! As I recall, the reason Bruce was after the bad guy was that the dude had messed up his sister. Bruce spent a lot of time first finding out who was responsible, fighting some minor bad guys, and chasing them around town on a motorcycle. Lets visualize how this whole process could be put into a game. It starts out with the main character helping his sister with something, like some guy trouble she is having. There is enough conversation and interaction for the player to get the idea he should like his sister and feels some attachment to her; wants to protect her. The player can also be gradually introduced to the play mechanics; gradually get a feel for the game. Enter the bad guys, who restrain the character while they mess up the sister and kidnap her, then they knock the character out. They leave some clues behind. The player must use these and other clues to find the badies, RPG style. He drives around town, VR style, perhaps optionally using the Arcade Racer. He finds himself in a warehouse, where he must combat some minor bad dudes, using VF style fighting and stick control, but with freedom to move around the warehouse. Multiple simultaneous attackers would be a nice added touch. Or perhaps allow them to be shot Virtua Cop style, with the gun controller. He then hops on a motorcycle and variously chases the bad dudes or runs from them, or whatever. He has a few other minor conflicts with the bad guys, must solve some more RPG style puzzles. He then stakes-out an office he discovered, and follows a bad guy to head quarters, on the bike or on foot. He fights his way up to the top using different fighting styles, beats the bad guy, and recovers his sister. In the final scene, they both watch the sun go down while sitting on the beach -- The End. Multiple paths to the ending could be employed. For example, if the player does not discover the motorcycle, he could get everywhere on foot, or in a cab -- the game adjusting to lead the player thru the story based on available character inventory and player choices.
As another example, I think it would be neat to combine VF style stick control with an RPG. Mages, thieves, fighters, and priests must deliver spells, attacks, and defenses using stick-and-button motions in real-time. Instead of using a fixed arena, the ability to render a lot of polygons is leveraged to allow the character freedom of movement; fighting takes place on the normal playing field, instead of in a different combat mode. As the character moves up in levels, more spells and moves are added, with either the player left to discover how to deliver them, or given training and hints. The environment could be simple visually, but use lots of motion to add realism. Trees and branches that sway in the wind, that rustle when someone is moving thru them. Non-player animals that hop and move around. Clouds and wind blown dust. Lightning, fog, and rain. Weapons and other helpful items that the player actually bends to pick up, instead of instantly acquiring. (In fact, in the case of locating weapons and items, flat shading may even help the player. I've been stumped more than once because I could not see an item due to the background visual noise.). NPC's that walk, run, and fight smoothly and realistically. I think you get the idea.
Perhaps I'm overstating the amount of time and memory that would be freed up by using simple flat shading. Perhaps all of this would be too demanding of the Saturn. However, with good object oriented design and data driven behavior, I believe based on my programming experience -- as limited in application to games that it is -- much, if not all, of this is obtainable.
In summary, I hope that I have made a good case for at least considering the use of flat shaded polygons to free up system resources, which are then put to interesting and entertaining uses; animating the world and otherwise providing a more engaging gaming environment.
James Hugard
internet: james.hugard@ast.com
voice: 714/727-8662
smail: AST Research, Inc.
16215 Alton Parkway
Irvine, CA 92718
mail stop 2-76
