Mastering the Game Boy Palette: How 4 Shades Became an Art Form

1. The Screen That Shouldn't Have Worked

When Nintendo released the Game Boy in April 1989, it didn't exactly look like the future of handheld gaming.

Atari had the Lynx, with a 16-bit custom graphics processor and a backlit color display. Sega was about to release the Game Gear, which could show 32 colors at once from a palette of 4,096. Then there was Nintendo's machine: an off-white plastic box with an 8-bit Sharp processor running at 4.19 MHz and a tiny 2.6-inch reflective LCD.

The screen could display only four shades, all somewhere between green and yellow.

On paper, it looked hopelessly outclassed.

It wasn't.

The Game Boy went on to sell more than 118 million units when the original system and Game Boy Color are counted together. The Lynx and Game Gear, despite having much more impressive display hardware, eventually became footnotes in handheld history.

There are obvious reasons for the Game Boy's success. Tetris certainly helped. Battery life mattered too. Four AA batteries could keep the machine running for around thirty hours, while the Lynx could burn through six batteries in less than four.

But hardware limitations had another, less obvious effect.

They changed the way artists worked.

With only four shades available, there wasn't much room for lazy shading or subtle gradients. Every pixel on that 160 × 144 screen had to do something useful. Artists couldn't simply add another color when an image looked flat. They had to find another way to create depth, texture, and contrast.

The limitation became part of the visual style.

2. The Strange Little Olive Palette

Open a modern Game Boy emulator and you'll often see the original graphics displayed in ordinary gray.

It looks perfectly fine on an OLED screen. It also doesn't quite look like a real Game Boy.

The original DMG-01, short for Dot Matrix Game, used a reflective Super-Twisted Nematic, or STN, LCD. There was no white backlight behind it. Instead, ambient light came through the display, reflected from the layer behind the liquid crystals, and traveled back toward the viewer.

That display technology gave the Game Boy its familiar olive-yellow appearance.

People sometimes compared the color to pea soup.

The four shades weren't simply four evenly spaced values between black and white, either. They came from the physical behavior of the LCD and the voltages applied to its cells.

The commonly represented palette looks roughly like this:

One thing stands out when you look at those values.

The jump between Shade 1 and Shade 2 is much larger than the jump between Shade 0 and Shade 1.

That gap was important.

Artists could treat the first two shades as a light pair: a highlight and a basic surface tone. The darker two could then establish shadows and edges. Moving from the light pair to the dark pair produced a strong visual boundary without needing a fifth color.

In other words, the palette wasn't particularly flexible.

But it was useful.

3. How the Game Boy Actually Stores Pixels

The Game Boy's graphics hardware was designed around the same basic idea: keep things simple enough that inexpensive hardware could move the pixels quickly.

Modern image formats usually keep the information for a pixel together. An RGBA pixel, for example, has four bytes representing red, green, blue, and alpha. Even indexed PNG images are fairly straightforward: each pixel stores an index that points to a color in a palette.

The Game Boy organized its graphics differently.

Its graphics data was stored as 8 × 8 pixel tiles in VRAM, covering the address range from $8000 to $97FF. Each tile occupied 16 bytes.

Those 16 bytes weren't arranged as a conventional stream of 2-bit pixels. Instead, the Game Boy used two bitplanes.

For each row of eight pixels, the first byte contained the lower bit of each pixel, while the second byte contained the upper bit. In other words, Byte 0 represented Bitplane 0 and Byte 1 represented Bitplane 1.

To work out the color ID of a particular pixel, the PPU took the corresponding bit from each byte and combined them:

That gives four possible values: 0, 1, 2, or 3.

At first glance, it seems strange to split the information this way. Why not just pack four 2-bit pixels into each byte?

The answer has a lot to do with the hardware Nintendo was working with.

The Game Boy used an 8-bit bus, so one byte could provide eight low bits in a single read. Another read supplied the eight high bits. The PPU could then feed those two bytes into separate 8-bit shift registers and shift the corresponding bits out as the display was being drawn.

This avoided the need for more complicated circuitry to unpack and rearrange pixel data continuously. For an inexpensive handheld built around an 8-bit processor, it was a neat compromise: simple memory organization, simple hardware, and enough throughput to keep the display moving.

4. The Palette Registers Were More Powerful Than They Looked

The color ID stored in a tile wasn't the final shade shown on the screen.

It was really just an instruction saying which of four palette entries a pixel belonged to. The palette register then decided which physical LCD shade that entry represented.

For background graphics, this was handled by the BGP register at $FF47.

The register was divided into four 2-bit fields:

• Bits 1–0 → physical shade for Color ID 00
• Bits 3–2 → physical shade for Color ID 01
• Bits 5–4 → physical shade for Color ID 10
• Bits 7–6 → physical shade for Color ID 11

The standard setting was %11100100, or $E4, which mapped color IDs 0 through 3 to the four normal LCD shades in order.

The useful part was that the palette didn't have to remain fixed.

A game could change the register while it was running. Since the PPU used the palette as it rendered the screen, changing the mapping could alter the appearance of existing graphics without rewriting the tile data sitting in VRAM.

That made certain screen effects surprisingly cheap.

Writing $00, for example, mapped every color ID to the lightest shade and could create an immediate white flash. Other values could reverse the apparent light and dark relationships. A sequence such as $E4 → $F9 → $FE → $FF could also produce a simple fade toward black.

Nothing in the tile graphics had to be redrawn for those effects. The same four color IDs were simply being interpreted differently.

Sprites had their own palette registers, $FF48 (OBP0) and $FF49 (OBP1), which gave developers another way to vary the appearance of objects on screen.

There was a catch, though.

Color ID 00 in a sprite was always transparent.

That meant a sprite could technically contain four color IDs, but only three of them produced visible shades. The fourth simply let whatever was behind the sprite show through.

For character artists, that was a meaningful restriction. A character such as Mario, Samus, or a Goomba had to be designed around three visible tones. The two sprite palette registers offered some flexibility by allowing different objects to use different mappings, but they couldn't change that underlying limitation.

5. The Visual Rules Artists Learned to Live With

Four shades on a reflective, unlit display don't give an artist much room to work with.

Over time, though, Game Boy artists developed a set of practical solutions. Some of them became so common that they eventually felt like part of the platform's visual language.

One of the simplest was the use of strong dark outlines.

The original Game Boy screen depended on ambient light, so the amount of contrast available to the player could change from one situation to another. A character that looked perfectly clear in bright daylight might become much harder to distinguish from its surroundings in a darker room.

A continuous outline helped keep the shape intact.

Games such as The Legend of Zelda: Link's Awakening and Super Mario Land 2 often surrounded characters with a solid dark border. It wasn't merely a stylistic preference. The outline created a clear boundary between the character and the scenery behind it.

Dithering was useful too, although it had to be used carefully.

The STN display was relatively slow to respond. When high-contrast patterns moved across the screen, the previous image could linger for a moment and produce visible ghosting. A checkerboard pattern might look perfectly acceptable on a stationary wall, tree, or patch of water, but the same pattern on a moving character could turn into a blurry mess.

For that reason, dithering was generally safer in static parts of the scene. Moving sprites tended to work better with larger areas of solid tone, where their shapes remained easy to recognize during movement.

Then there was the problem of highlights.

In a more conventional illustration, an artist might use one tone for the highlight, another for the main surface, another for the shadow, and a final dark tone for the outline. A Game Boy sprite simply didn't have enough visible shades to do all of that comfortably.

Something had to disappear.

Often, the brightest highlight was the easiest thing to sacrifice. Shade 1 could become the main body tone, Shade 2 could define the deeper shadow, and Shade 3 could provide the outline. The lightest shade could then be saved for a few small details where an extra bit of contrast really mattered—an eye, a buckle, or another tiny feature.

The result wasn't an attempt at realistic lighting.

It was more like visual editing: deciding which parts of the subject were essential and which could be left out.

6. The Screen Didn't Stay the Same

The original Game Boy hardware stayed around for a surprisingly long time, but the display itself didn't remain unchanged.

Nintendo gradually improved the machine and its screen before eventually moving the platform into full color.

1994 — Super Game Boy

The Super Game Boy was perhaps the most unusual step because it moved the Game Boy experience onto a television.

A Game Boy cartridge could be inserted into the Super Game Boy adapter and played through a Super Nintendo. On the television, developers could take advantage of custom four-color palettes for different areas of the game instead of being restricted to the original handheld's olive shades.

The system could also display colorful borders around the game screen.

The underlying Game Boy graphics were still based on four color IDs, but the television gave those four values a completely different visual range.

1996 — Game Boy Pocket

The Game Boy Pocket made the hardware smaller, but its display was an important improvement too.

Nintendo replaced the original STN panel with a film-compensated FSTN display. The familiar green tint was replaced by a much cleaner silver-gray appearance, and the improved panel also reduced some of the ghosting that players had become accustomed to on the original Game Boy.

It looked less like the machine people remembered from 1989.

It also looked considerably cleaner.

1998 — Game Boy Light

The Game Boy Light arrived in Japan with one feature that players had wanted for years: a built-in light.

Its electroluminescent backlight produced a distinctive cyan-teal glow. For the first time, playing a Game Boy in a completely dark room didn't require an external light or one of the clip-on accessories that had become common with the original system.

By then, however, the four-shade era was nearing its end.

The Game Boy Color was about to change the rules completely.

7. Converting Modern Images to Four Shades

A modern photograph contains far more tonal information than the Game Boy can display.

If you simply take a high-resolution photograph and assign every pixel to the nearest of the four Game Boy shades using ordinary RGB distance, the result is usually disappointing. Large parts of the image tend to collapse into the middle two tones, making the photograph look muddy and flat.

A better conversion starts before the image ever reaches the four-color palette.

Start by stretching the contrast

Photographs often contain a large number of pixels around the middle of the tonal range. One way to deal with this is to examine the luminance histogram and stretch it before quantization.

For example, the darkest 3–5% of pixels can be pushed toward Shade 3, while the brightest 3–5% are pushed toward Shade 0.

The exact percentages aren't universal, but the idea is straightforward: give all four shades some useful territory instead of letting most of the image pile up around the middle.

Use perceived brightness rather than raw RGB distance

RGB values don't correspond neatly to the way humans perceive brightness.

Our eyes are considerably more sensitive to green than blue, with red also contributing strongly to perceived luminance. A standard Rec. 601 conversion uses:

A perceptual color model such as Oklab can also be useful.

Either approach generally gives better tonal decisions than treating the three RGB channels as equally important when deciding which Game Boy shade is closest.

Pick a dithering method that fits the image

The final step is deciding what to do with the tonal information that doesn't fit into four discrete levels.

An 8 × 8 Bayer matrix produces a regular, ordered pattern. It has a distinctly mechanical appearance and can work well when the goal is an intentionally retro, highly structured pixel-art look.

Floyd–Steinberg error diffusion behaves differently. Instead of applying the same pattern everywhere, it spreads the error from one pixel into nearby pixels. That can preserve smoother transitions and is often more convincing for photographs, portraits, and other images where the tonal changes are gradual.

There is no single “correct” way to make a modern image look like a Game Boy image.

The important part is understanding what the original hardware could and couldn't do.

The artists working on the Game Boy didn't have enough colors to reproduce everything they saw. They had four shades, a small screen, a limited amount of memory, and a display that behaved differently depending on the light around it.

So they learned to simplify.

A shadow might replace several real-world tones. A handful of pixels could stand in for a face. A checkerboard could suggest a gradient that the hardware couldn't actually display.

What began as a hardware limitation eventually became a recognizable visual language.

And that may be the most interesting thing about the Game Boy's four-color screen: the artists didn't overcome the limitation. They learned how to make the limitation work for them.