Unit Appearance Diversity

[scienceguy8]

Also Known As: Sprite Artists' Worst Nightmare

The inspiration for this idea comes from two sources:
(1) Frankensteining: the art of cutting up Wesnoth Sprites and gluing them back together in a manner similar to Frankenstein's Monster.
(2) Left 4 Dead: a first person shooter developed by Valve Software, whose previous titles include the Half-Life series, Portal, Counter-Strike, and Team Fortress 2.

Valve included with L4D some information on the design process of the game built into a set of maps. There, the player can learn how enemies and player characters were developed, how music and color effects can be used to set the mood, etc. One Process they talked about really interested me, however, because it may be feasible to introduce a similar process in the Battle for Wesnoth.

Left 4 Dead's premise is that you are one of four survivors in a zombie-ridden environment and that you must work together in order to make your way to a safe house or extraction point at the end of the level, fighting your way through hundreds of sick people who all want to eat you. The problem is that in a single play session the four players can see and kill literally thousands of zombies, and if there is not enough variation between each zombie the game loses its believability, the player is no longer immersed in the game, and the game play experience feels more artificial. While Valve could have created thousands of different models for the Infected and scattered them about randomly, this would have increased the game's size dramatically and reduced its performance. Instead, they developed a system in which each infected starts with one of a couple different body models. Clothing textures are selected based on the current environment, a head texture is selected at random, skin tone is established randomly as well, etc. In this way, by using the resources used by, say, 25 characters, Valve could populate an environment with hundreds, perhaps thousands, of enemies with noticeable differences between them.

The idea is to use a similar procedure to create unit diversity in the Battle for Wesnoth. Each sprite would consist of a head, a torso, and legs. Bare skin would be defined in much the same way as team coloring is now. For each unit, there would be three to five variations of head, torso, and legs. When the unit is created, Wesnoth selects each component at random, selects a random skin tone based on a defined range in the unit's race file, and assembles and colors the unit in the game world. In addition, maps could be made to have a "temperature" tag that could be used to define that area's season and weather conditions. Based on this temperature tag, Wesnoth could change the odds on which components are selected, so in maps with snow terrain and a temperature tag set to low, units would be generated wearing winter gear.

Pros:
Increased polish to an already great game.
Increased believability of characters and events.
More potential characters for less HD space.

Cons:
Sprite creators would have to create at least nine times as many images as before, including animation images.
Increased program package size.
Imagine the bugs this system would have in the beginning.

[doofus-01]

Cons:
Sprite creators would have to create at least nine times as many images as before, including animation images.

I think that's a show-stopper. This idea would not reduce the workload by very much. And it's hard to imagine a way to automatically create animations that won't look frankensteined.

[Turuk]

And it's hard to imagine a way to automatically create animations that won't look frankensteined.

This is a very good point, since Wesnoth uses sprite art.


Your idea is one that, in theory, is indeed something that would add a great deal to the game. The army that a player builds will truly look like a diverse army, with two units rarely looking the same and even having the units change into the appropriate gear for their climate.

However, you are comparing a video game produced by a company that has the time, money, equipment, and, most importantly, large numbers of skilled workers with nothing to do but create that art. Wesnoth is a game that thrives on the contributions of users, who are sacrificing their free time in order to improve the game. While this sprite idea would certainly add a great deal in the way of diversity to the game, you have to be realistic and truly look if it was feasible or not.

How many good sprite artists are there now that could accomplish this? How many of those would be willing to work on this project? Would they be able to truly design the sprites so that every bit could be interchangeable without looking odd? And how much would this add to the game in terms of the engine working each scenario to formulate these units? Does it work offset the advantage of having varied units? Is this something worth coding or do other areas of the game have more pressing concerns?

[Lord_Aether]

I wonder if you could make the variations different enough for one unit from another... I imagine that trying to tell just what a particular unit is at a glance could become pretty difficult - L4D only had to make zombies, but Wesnoth has several different races, not to mention several different unit types within each race.

[thespaceinvader]

In theory, a fantastic idea. It's not the first time it's come up, by a LONG shot.

But in practice, virtually unworkable. Each new appearance multiplies the number of images needed, and all those images must be created. This wouldn't be hugely difficult for some basic things (adding a beard, for example) but it would be very time consuming.

That being said, I know one thing that Jetryl is planning (eventually) is to use an analogue of the TC system to introduce variations in skin colour, particularly for the orcs. That sort of thing is much more doable, since it requires only code, and a minimal amount of art changes.

[JW]

This would be an argument for Wesnoth to go 3d.

That plus the ease of making animations over 2d generally.

[faring]

I don't really think it is a good idea regardless of the difficulty involved in implementing it
Mostly because Lord_Aether's point
also you dont tend to have too many of the same unit at once in wesnoth whereas L4D has the zombie horde so it's not really that necessary
and how many variations can you get of a sprites face? :S I should imagine the armour for one type of unit would be a standard thing which is the same for all the others in the army so that wouldnt change and the colouring too would have to stay the same.. perhaps the only feasible thing would be skin tone but the soldiers would all be recruited from the same area.. unless maybe on campaigns if the map was a desert units recruited in that scenario could have a darker skin tone?

[turin]

Two words: Anti. Aliasing.

It's already impossible to anti-alias the tcoloring in with the rest of the sprite; do we really want to do this to skin coloring too?

[beetlenaut]

Sprite creators would have to create at least nine times as many images as before, including animation images.

Just to give you an idea of the scale of this project: There are currently over 2600 sprites! And each one represents hours of work. You are suggesting we make at least 20,000 more.

[scienceguy8]

And it's hard to imagine a way to automatically create animations that won't look frankensteined.

The changes would be purely cosmetic to each unit. No matter what combination of head, torso, legs, and skin tone, the animations would be exactly the same. Basically, to update each unit to the new system, each frame of animation would be separated into head, torso, and legs, changes would be made to each component to create new variations, and each component would then be called up during animation. The animations would not change, just the appearance of the characters performing the animations.

Two words: Anti. Aliasing.

It's already impossible to anti-alias the tcoloring in with the rest of the sprite; do we really want to do this to skin coloring too?

I don't know a whole lot about digital art, but have you tried layering? Create the unit as you normally would except the component that is to have variable shading. Create that using various degrees of alpha channel. Then when it comes time to color those areas render the color behind the unit so they are seen through the alpha.

I wonder if you could make the variations different enough for one unit from another... I imagine that trying to tell just what a particular unit is at a glance could become pretty difficult - L4D only had to make zombies, but Wesnoth has several different races, not to mention several different unit types within each race.

To simplify the animation process each unit, no matter how its appearance is rendered, would have the same pose and animation. The changes in costuming would also not be extremely significant. They would be different in appearance and ornamentation, but each variant would have the same color scheme and pose. One elven archer may wear a green long sleeve shirt, another with short sleeved shirt, yet another without sleeves but with a white long sleeved shirt under it, and perhaps another without the white shirt.

[LemonTea]

There are already some units with two sets of sprites for male and female variations, and they look almost the same as each other (I usually can't tell the difference unless I look VERY closely for a while). Which is a good thing, since if they looked wildly different I'd have to stop and think, oh wait, those are the same unit. So it's better if they look pretty much the same, for the purposes of quick unit identification.

When variations of the same units have to looks almost the same so that they're barely recognisable as different sprites, there's no point in drawing 10 times as many sprites as before, when players aren't really going to notice much.

[JW]

There are already some units with two sets of sprites for male and female variations, and they look almost the same as each other (I usually can't tell the difference unless I look VERY closely for a while). Which is a good thing, since if they looked wildly different I'd have to stop and think, oh wait, those are the same unit. So it's better if they look pretty much the same, for the purposes of quick unit identification.

When variations of the same units have to looks almost the same so that they're barely recognisable as different sprites, there's no point in drawing 10 times as many sprites as before, when players aren't really going to notice much.

I actually imagine that, other than skin tone variations, many minor base frame differences could utilize the animations already in place without seeming wrong.

[Turuk]

To simplify the animation process each unit, no matter how its appearance is rendered, would have the same pose and animation. The changes in costuming would also not be extremely significant. They would be different in appearance and ornamentation, but each variant would have the same color scheme and pose. One elven archer may wear a green long sleeve shirt, another with short sleeved shirt, yet another without sleeves but with a white long sleeved shirt under it, and perhaps another without the white shirt.

Right, but...

Just to give you an idea of the scale of this project: There are currently over 2600 sprites! And each one represents hours of work. You are suggesting we make at least 20,000 more.

I am pretty sure that no matter how else you would like to argue your point, this is the show stopper right there. We do not have many sprite artists as it stands now, and you are asking them to spend countless hours working on what is purely a cosmetic change.

2600 sprites * 9 variations (Your estimate) = 23,400

23,400 * 10 mins (Per variation) = 234,000 minutes or 3,900 hours or 162 days

Even if someone was incredibly skilled and could do each variation in a minute, that would take over 2 weeks of nonstop work. (16.25 days).

This is one of those ideas where no one is really saying that the idea is not a good one to add variation, but just proves to be entirely unfeasible.

[beetlenaut]

Basically, to update each unit to the new system, each frame of animation would be separated into head, torso, and legs...

That's actually quite a bit of work. Tedious work, too.

I don't know a whole lot about digital art, but have you tried layering?

It's not that we can't figure out how to do it, it's that the engine doesn't support it.

One elven archer may wear a green long sleeve shirt, another with short sleeved shirt, yet another without sleeves but with a white long sleeved shirt under it, and perhaps another without the white shirt.

I think it is time for you to put your money where your mouth is. Make some variations of a unit or two, and split up the sprites into pieces. Maybe some dev will feel like playing with the pieces if you've made them. If you don't feel you can do that, maybe you could take a programming class so you can work on the engine yourself (assuming you don't already know C++). If you can't do that, maybe you could start a fund to pay one of our artists. Something.

Sure, it would be nice to see variations on the units, and a better AI, and the ability to write campaigns in your favorite language, and a dozen other things. Everyone has a suggestion, but only a few are willing to do the work it takes to implement it. Show me you are willing to do some work instead of just suggesting that others do it, and I'll be less grumpy.

[scienceguy8]

Sure, it would be nice to see variations on the units, and a better AI, and the ability to write campaigns in your favorite language, and a dozen other things. Everyone has a suggestion, but only a few are willing to do the work it takes to implement it. Show me you are willing to do some work instead of just suggesting that others do it, and I'll be less grumpy.

I got a semester of basic C++ under my belt and am now downloading the development version source code as we speak. Let's see what I can do.