Pixel a Capcom Background
Backgrounds are something spriters don’t often think about pixeling. We see them plenty of times in games, but then just assume they use special programs for that sort of thing. Perhaps, but that doesn’t mean you can’t do it yourself, and in a reasonable amount of time.
For the adventurous of you want to give this a shot, you should know that pixeling a full background will take some trial-and-error, and a decent amount of art skill. As a rule, the art has to look good before the pixeling to look good after the pixeling.
Does that mean the best and safest way to make a good background is pixel-by-pixel? Good lord no. You wouldn’t draw a character by starting with its outlines – you’d start with a solid foundation and draw simple base stick figures and measurements to make sure you built everything in proportion, and then build up.
In the same way, instead of pixeling from nothing (where you’re only guessing where the next pixel should go and you have no idea what the final image will look like), wouldn’t it save time and give you more confidence to draw out the whole background first and then convert it into pixel art that you only have to clean?
–And that’s my big pixeling secret that’ll probably get me in trouble with the other hard-working pixel artists out there who do things the long way: I get my best pixel art when my pixels are cleaning, never when they’re building. Make that your motto and you’ll fool a lot of people into thinking you’ve spent a lot of time practicing with a lot of pixels.
But how do you draw a ‘base stick figure’ for something this big? It takes some work. But there are some shortcuts in Photoshop that’ll save you time, keep things accurate, and even keep your color palette under control. I’ve also been reminded that lots of people use Gimp instead of PS, so I’ve tried to include equivalent Gimp methods, too.
So let’s take a look at making backgrounds by going through my own learning experience when I did the Cats Lair.
STEP 1: Getting references

References are more important than normal for backgrounds, because you need not only the details but believable lighting.
I definitely needed references for this specific stage. And the more the better, since the Cats Lair has often been drawn in slightly different ways. I found as many pics of the same angle as possible, picked out the best parts of each pic and made a collage. I also kept pics on the side that had the best lighting and colors.
I also picked up another set of references: existing game backgrounds that I thought matched the color palette or did shading/dithering in a way I needed. I’ll explain why soon.
References are useful for getting an idea of several important things: general color sense, for the palette; the perspective the whole stage will need; and since this was for a mugen stage, it helped define the proportion for where the characters would stand, how high and how wide everything would have to be (adjusting for the lifebars that would be on the screen, etc).
When it comes to image size, you can choose to work at 200% and shrink everything down at step 6, but it’s not vital.
References would also help if you’re doing your own made-up stage, since consistent lighting and colors are difficult to get with a whole background on your own. Colors and lighting should be as believable as possible. So if you’re doing a castle you thought up, look up photos of real castles and professional artwork of illustrated castles. The majority of the details in your references may be ordinary, but they’ll be a guiding light towards believability.
STEP 2: The lineart
With the references arranged right, I was ready to start tracing lines. This is where different people can try different things – maybe you don’t want your background to have linework (an open field, a painting-like nature scene, etc) so you could skip this step.
I did the tracing in Illustrator, but any vector program would work (there are dozens of good free vector programs out there). A vector program is an extremely good idea for the lineart because the lines you make are completely and easily adjustable if you make a perspective error, or find any major adjustment you have to make. You can resize part or all of the linework, and stretch out whatever needs widening. And when you’re all done, you tell the program to stroke the lines with a width of 1 pixel, and you automatically have a perfect single-pixel outline.
The Cats Lair was traced from reference, but the linework seemed too narrow. I wanted something thicker, with more presence. Since everything was still in vector format and I wasn’t committed to any pixels yet, I was able to stretch out everything under the head in three mouse clicks, while selecting a few anchor points at the top of the bridge and dragging them closer together to easily solve a perspective problem that otherwise would’ve required a redraw.
Once you’re done, save all your paths. At this point, whether you continue to use the vector program or move to Photoshop/Gimp (“PS/G”, from now on) is up to you.
STEP 3: Quick Mock Shading
This is a planning step that will give you a customized blueprint for consistent, believable lighting. No need to be exact here, since it’ll just be a guideline, but don’t leave out much. Simply block in where the light will hit and where the shadows will be, and roughly establish your contrast. Just work in grayscale for now.
I put on a basic top-middlish lightsource for everything, but at a slight angle. I gave the head and claws some real contrast since they were metal, and I gave myself an idea of how I’d want to do the main stone body and the shoulder areas.
The main test: if you zoom out and glance at it, does it look roughly 3D?
STEP 4: Color palette decision time

*Stolen!* (with a blue hue shift thanks to Photoshop)
This may seem the most simple step. It is not. Getting all the colors to work together is a whole course in itself, but there is a shortcut: don’t be afraid to steal a color palette from other backgrounds. Doing a white building? Find a background that uses a range of whites for something. Getting exactly the right colors now will save you a lot of work and decision-making later when you’re applying those colors. Be sure and separate each group of colors: I kept the whites, metal blues, rock browns, grass greens and sky blues as separate palettes.
If the colors you find are close but not quite what you want, minor adjustments in PS/G’s color-correction tools will do the trick.
–Before we go any further!
It’s time to show you the whole basis for my auto-dithering process. This technique is so simple it’s stupid. Normally, I color a sprite in PS by doing brushwork under enlarged pixeled linework in RGB mode, shrinking it to pixel size and converting it to a specific color palette by converting from RGB to Index color (and while in Index, selecting the .act file with my palette, then choosing ‘Dither: None’), then cleaning a little. (See my Speed-Spriting Tutorial for the whole process)

The gradient to be converted into 50% dithered pixels
I wanted PS/G to also dither my colors now, but only at 50% (so the pixels make a checkerboard pattern). The closest I could get in PS was Pattern dither, and Gimp is Position dither. I had to do something to my colors beforehand, somehow.

The pattern dither: too messy and over-pixeled
Here’s a simplified look at how the trick works. You’ll want only a “50%” color dither on you final image, so first, via Index Mode, reduce the image’s colors only to the ones you want, as well as the 50% blend between each consecutive color.
You’re now in index mode. Go back to RGB mode so you can repeat this step one more time, but now reduce those colors to only the colors you want – using a pattern/position dither.

The result
Boom. Right there’s instantly the exact dithering I want, where I want it, and I didn’t have to touch a pixel. It was right here that I knew I could make this stage.
The one catch is that you have to have your colors in the right order before you make their 50% blends. To recap, here’s how to make a Gameboy background without moving a single pixel:
(Note: a Gameboy background is the perfect way to start practicing this tutorial)
STEP 5: The gradient tool – your new best friend
Here’s where you could choose to use either your vector program or PS/G. Some people, once they get used to Adobe Illustrator’s method for doing gradients, can’t live without it. You definitely get more options on how to work your gradients there, but really all you need in raster programs is a lasso tool and magic wand to control where your gradients go.
First, fill in all the base colors at their darkest shade. Separate each grouping of color palettes to different layers (you’ll see why soon), so in this case all my white parts (arms, body, lower area, floor) were on one layer, all the metal blues (head, claws, bridge, arm plates, door) were on a different layer, and so on.
Then begin adding the lighting with gradients. Get as natural a look as possible, and try to match the lighting from your references. This is the part where I mentioned you should be a decent artist. It doesn’t have to be fancy, but try to get some accurate lasso shapes for the more rounded parts. This is also where a good palette will pay off – slipping some of the greens into the white areas helped keep things from getting too boring and plain, and since the colors were already in my palette, I wouldn’t have to adjust anything afterwards, just slap on the gradients and let the upcoming color conversion do its work.
Other areas like the claws didn’t need too much gradation, but colors did have to blend smoothly in those reflections. So I went to a solid brush and put down some tones, then used the smudge tool along the edges of the colors to lightly blend a little. Too much smudging, and the dithering would spread out too widely. Excessive dithering can be distracting if your background is complicated.
A special note on non-solid objects (like the sky, in this case): resist the temptation to overuse low opacity brushes to build up color and volume. A little bit is okay, but you risk an over-dithered look. Use a solid brush, block in solid shades, refine, then smudge out.
After taking care of all the major areas, I found certain parts needing more finer-detail work, including the doors, steps and claw details. They would have to be pixeled after everything was converted, so I held off on the smaller details for now.
STEP 6: Color tables and index color – an hour of work in 20 seconds
I had the final colors I wanted this to be, and now it was time to convert everything into those colors. First I had to turn these colors into a palette that PS/G could work with.
I needed two palettes for each set of colors – one with only the colors I’d end up with, and one with those plus 50% blends of each. So for the whites, I opened a new PS document and penciled in each color of the white palette I was using, making sure to keep them in the same ‘brightest to darkest’ order that I was using in the pic. Then to save that as an .act file, it’s Mode > Index > Exact (with no dither), then Mode > Color Table > Save. I called it Whites_Single.
You do this with Gimp differently: First off, Gimp doesn’t use .act files. You just save your palette in the Palette dialog/toolbox. You also copy colors from your image to the palette differently. So with Gimp, I would’ve brought my colors into a new document, gone to the Palette toolbox and right-clicked somewhere inside to get “import palette”, Select Source: Image, Palette Name: Whites_Single, made sure the preview had what looked like the right colors, and hit Import.
Next, I had to double the palette to get the 50% shades in there. This simply required duplicating the layer, moving it over a few pixels and setting the layer opacity to 50%, then cutting off any overhanging parts. Then the Mode > Index (or Palette toolbox > Import, for Gimp) trick was repeated, and I called it Whites_Double.
Now it would pay off to have each color set on its own layer. I simply took the white’s layer and dragged it in a new document. Then I converted it with Mode > Index > Palette: Custom (load Whites_Double.act), no dither. Quickly did a Mode > RGB, and then finally Mode > Index >Palette: Custom (now load Whites_Single.act) with dither: pattern.
In Gimp, I would’ve taken the white’s layer to a new document and hit Image > Mode > Indexed. In the dialogue, then chosen Use Custom Palette (then choosing the Whites_Double), with the dither option of None. Then I’d get it back to Mode > RGB, and repeat the Mode > Index, but this time choosing Whites_Single for the palette, with Dither: Positioned.
*A little side note, although this does a pretty decent job of copying the PS technique, it isn’t quite perfect, so you might have to clean up some pixels here and there that don’t quite look right.
Not bad at all.
Drag that back to the original file and repeat with the other elements.
STEP 7: The finer details
One thing I didn’t think out all the way was how to best handle the original linework I traced. Since I had them on their own layer on top of everything, I locked the layer transparency and filled in everything with their surrounding color, but then I had to blend them in at some of the edges. What I should have done is saved the original paths I made, so that I could now fill in the pixel lines with the surrounding color, then on a new path stroke those same lines’ paths with a good outline color with a 2px brush. Then I could re-apply the Index Color trick to reduce the blurred brush lines to almost perfectly anti-aliased lines.
That trick of re-applying the Index Color trick came in handy as I was doing the lines for the shield plates on the first two stories and the details on the claws. Simply using the line tool set to 1px with antialias turned on, I put down the lines, which were fuzzy at this stage, and those window slots. Then I just brought them to my set color palette with Index color and voila, no need to do antialiasing by hand, except to clean up what didn’t quite look smooth in my 100% thumbnail view I had on the side. Always have a 100% thumbnail view of what you’re looking at, on the side.

1 Before 2 The line tool adds edges and edge highlights 3 The color palette is reduced
Next, there are certain pixel art conventions you can use to make details stand out more. One of them is adding a highlight to segmented areas in a wall (see the pic).
I also found there were some areas that my gradients didn’t handle quite right, so I’d have to dither parts of this by hand. There are two very helpful ways to speed up the hand-dithering process: making a dither-brush and making a dither pattern-fill:
Finally, since I separated everything by color palette, I saw I needed to do a little blending of certain elements into the background, like the mountaintops into the sky. Since I don’t want to add more colors to the palette, I worked on just the mountain layer, and went over the existing edges with little bits of its brightest color, to fade the dark rocky area’s edges into the brighter sky. Just a few touches of the lightest rock color on it’s sharpest pixel curves was enough.
STEP 8: Finding your own methods
The entire Cats Lair – minus effects and cameos – came out, amazingly, to only 111 colors. It may not be exactly how Capcom would’ve done a “ThunderCats vs Street Fighter” stage.. but then again, I didn’t have to earn a paycheck with this.
I admit this isn’t the cleanest and simplest way of doing everything, since this was my first time trying backgrounds. You can definitely refine these steps further. For instance, in PS, you don’t always have to save an .act file every time you have a new palette: just put your colors in a new document with the pencil tool, convert it to index using ‘exact’ colors, then move to the document you want converted to that palette, and then convert it to index using ‘previous’ colors.
Perhaps you’ll discover your own way of doing other parts, or other kinds of backgrounds. But no matter what techniques you use, the main tricks in this tutorial will definitely help you save time, preserve your eyesight, and make you the envy of harder-working pixel artists everywhere in your quest to dither the perfect background.
Cool Tutorial, but I’m having a problem with making a Dithering brush for GIMP. I followed what you said but when I go to draw with the brush it over laps and fills pixels it’s not suppose to. Ive used a tablet and the mouse to draw and i have the same problem. Any ideas?
Take a look at the “spacing” in the Brushes menu and set it to 100. When you move your brush, it’ll space out each mark by 100% which means in this case it’ll look like a continuous dither pattern as you move your brush along.
Nice tutorial, dude. One question though: Did
Capcom use this method when making backgrounds
for the Street Fighter 3 games?
I don’t know, but they probably used at least a little bit of ‘index painting’ here and there, judging by the not perfect dithering that gets used for some gradients. I still think they had specialized software for some of their art.
This is a nice tutorial, but I got stuck in step 4. How do you manage to get the 50% blend between each color?
That is the trickiest part of the whole technique, but once you understand each step involved and why you have to do them, it becomes really simple:
You have to choose ahead of time what your palette will be. Do that by either stealing colors from an existing game background or by choosing some experiments.
Once you decide on those colors, make a new document in photoshop and add those colors in, then save them as a .ACT file. That’ll be your first .act file.
Then take those colors and arrange them from darkest to lightest. Create a mid-color for each consecutive color so that you have a 50% blend of each of them (see the gameboy example of ColorsDouble.act). Save that as your second .ACT file.
An .act file tells photoshop that its only allowed to use those specific colors in an image. So you can take a piece of artwork, tell photoshop to convert it to Index Mode using your second .act file, and that’ll cut the colors down to your palette plus the 50% colors you made. Then you convert it back to RGB Mode so you can do the trick again, but now you’ll convert to Index mode using your first .act without the 50% colors, and this time you’ll tell it to dither out any missing colors which means it’ll dither out the 50% colors. And since those missing colors were 50% mixtures of your palette, the dithering that photoshop will do to them will be that beautiful checkerboard type.
Oh, it’s much clear to me now. I tried to create the effect, but i did something wrong: some areas in the image get the checkboard dither, but some others don’t.
I have some question, i hope you can help me again.
1. Must the colors in the first .act file be arranged in some way? In the game boy example, they are ordered from lightest to darkest
2. The order of colors in the second .act file is darkest to lightest or lightest to darkest?
Thank you very much, kiwi. I’m new to your page, and all the tutorials are very useful. Sorry for my english
You put the colors in order, either darkest to lightest or lightest to darkest, only for your own convenience (an ACT file doesn’t record where you put your colors, but only what colors there are).
So if you were doing a gameboy palette, you would first use the pencil tool to put a spot of the darkest, then next-lightest, then almost-lightest, then lightest colors, and have them all next to each other and touching. You stop for a second to save that as one ACT file, but the reason for putting the colors in that order is only so that you can have an easy way of making the 50% color blend: by using the pencil tool set at 50% opacity and taking the darkest color, you can put a tiny amount of that into the next-darkest color and thus make a blend between them. Then you repeat the steps by taking the next-darkest color and putting 50% of it into the almost-lightest color, and so on. Because you put the colors in order, you can now keep track of which colors you’ve blended as you move through each one.
You have to make sure you’re using exactly 50% opacity, or else the dithering won’t always be a perfect checkerboard pattern.