Thread by @BryanRenno

Theres a unique skill in the history of games that was very brief but I've always been fascinated by: the transfer of 3D renders to sprites. So let's take a look at one of the most famous examples: Donkey Kong Country. How did Rare transfer this render to a 16 colored sprite?

You might look at this now and go "just scale the render down to a lower resolution and you're done!" But that isnt something you could do with the SNES hardware.

Scaling the image is one thing but the real problem was color limits on sprites. 16 colors is was the hardware limit

This is fine when you worked with the more common way if making sprites but when Rare began exploring the use of 3D they ran into a big hurdle, the rendered images had millions of colors. Different shading, textured surface and lighting combinations. Just look st Dixie's hair

Each shaded corner is blending evenly. You could not do that with this limit. Just take a look at how theres an implied sense of shading but it's just different colored blocks.

This is zoomed in but at the scale most players will see this at it blends in nicely

So Rare had a problem: how do you adjust this? Well at first they decided to do this by hand. Scaling the sprite to either a 8x8, 16x16. And then it was hand editing pixels on top of the rendered image. This was fine for static backgrounds but not animated characters

They were more concerned about the consistency of the colors when a character was animating. The compression of each render adjusted colors on each frame differently

But in practice, because in motion you dont notice the subtle color changes on each frame, it wasnt an issue.

Ultimately Rare developed their own automation tool for editing the color palette of rendered characters that sped the process up tremendously.

This was great for characters but backgrounds were a different story. They had their own sets of problems

They had to brute force their way. Each background element had to be rendered and converted to smaller tile sets. Imagine breaking up ground tiles into multiple tiny squares and making sure they all lined up with multiple variants of other broken up tiles.

All by hand

And background tile atlases had to share limited color palettes as well. One of the crazier tricks they had to pull off is having every interactable prop (crates, barrels, bananas, coins, letters, etc) had to share the same palette of 16 colors! The same shade of brown

Theres an art to preserving the look of 3D in these heavily compressed ways but still making the art flexible enough to use in different ways that the art team at Rare did a great job with. These techniques would be applied to other SNES games and in the following generation

As hardware got better at rendering real time 3d with more complexity and at higher resolutions, the art of making pre-rendered content like this has disappeared for the most part from games.

In some ways, it's a shame to see skills and art work like this represented gone.

Big takeaway from this is how much work it takes to get art to the quality bar you want, even with all the tools, automation and happy accidents that happen along the way. Even small things like what color one pixel is changes the whole look. And that takes time to balance.

If you eant to learn more about this I recommend two videos:

One is probably what made me seriously consider making video games as a child. It has some interesting bits but is mostly 90s 'tude.

The other is @digitalfoundry's episode talking to the devs of the DKC series. This one is fantastic and goes over a lot of the same points I did in greater detail.

And if you want to see some great examples of pre-rendered 16 bit games I'd recommend the DKC series (obv), Killer Instinct, Rendering Ranger R2, Toy Story and Blackthorne just to name a few.

Woah this blew up. All I have to promote are my environment art mentorships for game art.

https://twitter.com/BryanRenno/status/1150781730543898630?s=19