These Photos Contain Exactly One Pixel of Each of the 16 Million RGB Colors


For you and me, RGB color spaces may just be an obscure but important mechanism towards achieving properly color-balanced photos. For a certain group of image nerds, however, it’s the whole enchilada.

Welcome to the allRGB Project, an ongoing effort that challenges digital artists/programmers to create images that use each of the approximately 16 million colors that comprise the RGB spectrum.

Before we get to the photos, a little bit of science is in order. If you’d like to skip the science, we promise not to judge you… just gloss on over the following paragraph.

The RGB color model posits that all colors are mixtures of three primary hues — red, green and blue — which can be combined in various proportions to produce 16,777,216 distinct colors discernible to the human eye. Various specifications based on the RGB model allow consistency among display screens and digital images.








The rules are pretty simple: Use whatever algorithms or tools you like to create an image of 16.78 million pixels, one for each color in the RGB spectrum. Each color may be used once and only once.

So far, the project has collected 99 entries from 36 artists, ranging from geometric abstractions to recognizable images with striking coloration. GitHub has a free tool for automatically generating such images, or you can do your own math.

(via Pop Photo)

Bonus weenie tip: Load one of these images into Photoshop and check out the histogram.

Image credits: Space Shuttle by Benny, Barcelona by Eric Burrnett, High Quality Render of Wedding Photos by 1COmMJz2, Naamloos-1 by sjoerd, Flowers by brandf, Reddit by brandf, M.I by ACJ and Chilly Run by Death9. All provided by allRGB

  • russianbox

    This doesn’t seem to work, on the launch photo i can find 5 of the same pixel

  • Guest

    I mean 5

  • Tom Williams

    You have to look at the photos on the web site, not these resized copies. To have one pixel of each color the photo would need to be large enough to contain 16,777,216 pixels – a square image would be 4096×4096 pixels.

  • zeptom

    This only work on the original images on there site, scaling them down like images in this article and you lose pixels or get more then one of the same…

  • Ralph Hightower


  • Jeffrey Friedl

    “RGB color space”? “Properly color-balanced photos”? “The 16 million colors that comprise the RGB spectrum”? “Distinct colors discernible to the human eye”?

    I know what you wanted to say, but wow, the technical inaccuracies here are impressive. “Look at the pretty pictures; each pixel is different color!” would suffice to get the same point across without slinging technical terms in factually-inaccurate ways.

  • Zos Xavius

    Indeed. The BS level is deep with this one. First of all there is no standard rgb color space outside of sRGB which is a generally accepted standard. There are quite a few RGB colorspaces. Secondly there is nothing in rgb’s “specs” that determine color “consistency” (that’s a laugh right there!) across displays! Its a color space! It defines a range of colors. A given display may or may not even be capable of producing the gamut required to reproduce all of sRGB. 16.7 milliion is the number of colors possible under 24-bit (8-bits per channel) RGB. Any photographer that works with RAW files should know they are decoded into 48-bit files (16bits per channel AKA 16-bit RGB in PS). I would have hoped that someone writing for petapixel would know the difference between bit depths and color space. Pretty basic stuff in the world of digital photography. Wikipedia has some great pages on RGB and colorspace.

  • Chris Pickrell

    Wasn’t sRGB created to create a standard for all monitors and browsers?

  • Jeffrey Friedl

    It’s even worse (and simpler) than that… clearly the site is speaking of 24-bit RGB (8 bits per Red/Green/Blue channel) as you mention, but this concept is completely unrelated to the concept of “color space”.

    Each pixel is represented by three numbers in the range of 0-255, but exactly how those numbers map to colors is not defined in the absence of a color space; any particular submission could different color spaces applied to yield different looks, all without changing the every-pixel-is-a-different-color correctness of the submission.

    This is all tangential to the “this is quirky fun” nature of the project, but the concepts are important to this blogs’ demographics, so it’s prudent to clarify the bigger mistakes, lest readers as unfamiliar with them as the original author fall in to the trap of “learning” from the article.

  • Brian Cooper

    Wow, kind of trippy!

  • Jeffrey Friedl

    sRGB is a color space (a mapping from pixel numbers to device-independent color) created in the late 90s to approximate the common consumer CRT of the time, and as such, if things that produced digital images (like cameras and Photoshop users) chose the sRGB digital representation for the image color data, common folks (consumer monitor, software that wasn’t color managed) trying to view them would likely get colors that were sort of close-ish, at least within the randomness imparted by the user’s fiddling with the CRT’s knobs (“tint”, etc.).

    Digital-image color data are just numbers. The “color space” is the scale that the numbers are to be interpreted in.

    If you get an image that doesn’t indicate which color space should be used to interpret its data, in the absence of other clues, guessing sRGB is often your best bet, because of sRGB’s history. But otherwise, there’s nothing particularly special about sRGB. It’s the most restrictive (least expressive) of commonly-used color spaces.

    I wrote a primer about this some years ago:

    Digital Image Color Spaces

  • Sort it out


  • zeptom

    Thanks for pointing that out. English is not my native language.

  • kassim


  • Peter Neill

    also, the moment its saved as a jpeg, no matter what the size you will begin to loose the effect as even at max quality in jpg, the compression system will blend some pixels that are almost the same colour if they are neighbouring pixels

  • Final_Word


  • Lauren

    Why not?