Hacker News Reader

Comment by onion2k | Hacker News Reader

I honestly thought pixel art style games were made by creating 3D assets and then rendering them in low resolutions as sprite sheets these days. Or using the 3D models in the game with a 'pixelate' shader. The idea that people still hand draw sprites slightly blows my mind.

View on HN

Dead Cells does the "2D sprites from 3D models" thing.

It requires some specific models and techniques that wouldn't look good in 3D, though, it's not just cell shading.

But it lets them have a wild variety of player skins and enemy animations without having to redraw everything by hand.

Why? It's much easier to draw a pixel sprite instead of creating a 3d model and then doing tricks to render it as 2d..

And hand-drawn images look better if the artist is skilled.

King of Fighter 12 sprites, which are regarded as being high quality, were made from 3d models. They first animated 3D models, selected specifics frames, then traced sprite over those.

https://kofaniv.snk-corp.co.jp/english/info/15th_anniv/2d_do...

> then traced sprite over those.

In the past this was called roto scoping and used to capture outlines and movement of real objects for animation. The end result is still a hand drawn and shaded object instead of just a screenshot of a posed 3D model.

ah yes. now that you say it, i recall they did that for Terminator 2.

https://www.youtube.com/watch?v=D0xp74uIZO4

Why? It's much easier to draw a pixel sprite instead of creating a 3d model and then doing tricks to render it as 2d..

It's never a sprite though. Even back on the Super Nintendo game sprites had 9 angles * lots of actions * lots of frames for every animation. Multiply that by different lighting in modern games, and using the power of a 3D engine starts looking like an obvious choice. I know lots of games do this for environments. I assumed they did it for everything.

3D doesn’t transfer well to stylized 2D. That’s one of the primary reasons why people are still doing it manually. Pixel art also has a much lower barrier to create.

It's a very high barrier. Because you can't interpolate animation, or move bones around, or slap a different texture if you don't like something.

Every motion in every direction has to be animated. It's akin to creating animation by drawing every change on a separate sheet of paper, photographing it, and then stitching together.

The barrier of entry is indeed low. It doesn't take much to start sketching some pixel images. Making a coherent animated whole out if it? Very labor intensive and hard.

During the 80s the pixelized sprites were drawn to be as realistic as possible given the inherent limitations, but when you make a retro-game now, you probably want to target the de-facto aesthetic itself, and then my guess is you don't want to go too far in the realism direction. Clever use of few animation frames and angles conveys this better than blindly just rendering and scaling down using some optimal resolution reduction algo..

I think (but I am not sure) the sprites in Fallout I (1997) and Fallout II (1998) were created as re-touched 3D renderings.

(I am instead sure they revealed that the characters in video animation were re-touched recordings of created puppets.)

Many techniques may be part of the creative process and of the production.

Factorio's graphics are 3D rendered via a pipeline to 2D sprites, multiple angles for the character & vehicle models, but only one for buildings since the camera angle is fixed. I mean that's not really pixel art anymore (fairly high resolution), but still a 3d-to-2d pipe.

Only if you keep it really low resolution and low framerate. If you want something higher resolution, multiple camera angles or fluidly animated doing it in 2D quickly becomes impractical.

Guilty Gears for example started out with 2D sprites and then went 3D with Guilty Gears Xrd[1][2], but with a lot of trickery to keep it looking 2D'ish, though that always was more anime-style than pixel art.

[1] https://www.youtube.com/watch?v=CTUvSsOtPKA

[2] https://www.youtube.com/watch?v=yhGjCzxJV3E

Despite what others have replied, you're right. Both of those techniques are used a lot. Still plenty of people doing it by hand, but you can achieve some really interesting results by "faking it".

If you can draw well(like, if you've spent a few hundred hours trying to carefully and accurately copy stuff), it's easy to outline the major proportions of a posed character or object quickly by looking at a reference and modifying it a bit. And if you have a palette set up as well, the rendering can go really fast once you know how you're going to group up the pixels(which is a major artistic decision, since clusters of same color pixels tend to read better as shapes and textures). So the payoff for 3D tends to manifest only after you have a lot of frames to render - each frame, on its own, tends to be fast to paint.

Basically, if you look at where CG is used in contemporary anime, that's where the line is drawn. Many shows will CG their vehicle shots and parts of their action scenes since they call for a lot of perspective drawing with swooping camera movements. They may use and trace over CG dolls to do establishing shots or get static posed characters in a scene, but once detailed acting is called for they tend to revert to drawing in keyframes. Again, generally relying on reference footage to get the acting down, but modifying it to fit the character designs.

There are some games that do as you say, "fake" some pixel art using a shader on 3D models. A Short Hike and Fez come to mind as examples.

A Short Hike isn't pixel art though, is it? It's too high-res.

And what blows my mind is that you actually think people are doing this. Sounds like the hardest and most convoluted way to do it. Scaling a high def image down to low resolution doesn't automatically create pixel art. It creates an ugly mess that would require as much work to look good as if you just did it in that resolution directly.

This guy uses the technique a lot with fantastic results: https://youtu.be/KPoeNZZ6H4s?t=861

While it looks great it feels very different from pixel art. More like a pixelated filter.

I think this is a better video from him showing how it looks in practice: https://www.youtube.com/watch?v=1FrIBkuq0ZI&t=420s

Reply on news.ycombinator.com