[Team & Technical] Approaching the decision of art style
Nov 19, 2018 22:51:16 GMT
disastercube and Galda like this
Post by ndrs on Nov 19, 2018 22:51:16 GMT
Approaching the decision of art style
Hi there all!
I want to make this thread to discuss a few problems we have with choosing our art style for the game. This concern came to me when seeing how we can and are progressing about general sprites for the game.
The main objective of this thread is to give a little insight on 2 methods on making in-game art style, so you are informed and can take a decision, or even try to learn and make a sprite!
For now, forgetting technical details we have 3 main lines of asset contribution:
- Lore/Design independent assets. Free to do: includes things not bound to world building or lore, like terrains, props, containers, generic objects, etc.
- Lore/Design dependent assets. Bound to Lore and world building progression: Things like buildings (specific to cultures), characters, spell animations, etc.
- UI Sprites: Some of them have dependency on lore/design (items icons, weapons...)
We currently have 2 problems regarding those lines of contribution
- Art style decision restriction: We don't have a final art style decided.
- Team technical capabilities: We don't really know how many members of our team have the capabilities for the art style decision
And here is where I introduce my main concern:
Right now the main agreement with our art style direction is to use 3D Pre-rendered sprites. That is technically costly to do, and we have few members of the team capable on doing this with viable quality.
Based on this concern, I want to explain 3 things and conclude with my personal opinion, so we can discuss this a bit.
1. Art style choices: Explanation and examples.
To understand the problems that this choice implies we have to take a look on how each of the choices work and a few examples.
Before starting, I want to state which choices are being considered, from my point of view:
- Pixel Art
- Pre-rendered 3D models
I'm making this choice assumption based on the discussion had on this thread. At the start of the project we kind of agreed on the game being 2D. 3D takes into account a lot of problems engine-wise, and all of that problems would have to be patched and adapted to make the game feel 2D. So, in summary, we wanted to have a game that feels like Golden Sun 1&2. We just were discussing about faking 2D in a 3D engine was better than just using 2D.
About this last topic I have to say, being in an era where indie games are proliferating, and 2D being the easiest way to one-person work or even small-team work, engines like Unity are built to have a lot of easiness working with 2D games. So faking 2D via 3D engine hacking doesn't seem like a good idea.
These 2 art style choices were selected because, once 3D options were removed from the thread discussions, we had 6 comments supporting Pre-rendered 3D, and pixel art being the second more referenced term on the thread in contradistinction to the difficulty of 3D modeling / rendering. Also, there was statements around the idea of mixing both two, in an effort to unload some technical weight on the team.
So, let's begin with some explanation and examples of both.
1.1. Pixel Art
Pixel art is the first raster based method of drawing elements on a game screen, following vector graphics.
Sprites started being just color matrices, like in Super Mario for the NES, where storing color pallet references on a matrix was the most efficient way of fitting the game into the cartridge. This means pixel art started as being inherently what's called pixel perfect, i.e each pixel of the sprite aligns and gets rendered on one pixel of the screen. Here's a little example.
![](https://cdn-images-1.medium.com/fit/t/1600/480/1*myrco8Z6xF0wwD_u7YJrgA.png)
At the left, interpolated sprites.
A the right, pixel perfect sprite matching.
Now a days, most games that use 2D pixel art don't render sprites at pixel level anymore. They just use no interpolation but pixel boundary calculations and camera orthographic size / pixel-per-unit adjustment. This technical gibberish comes to say this: Pixel Art means boxy 2D art, as 2018.
Also, pixel art has no size restrictions anymore, and it's just a way of making lines and shape more clear. This means nowadays we can draw really cool sprite art animations, like for example this animation:
![](https://i.gifer.com/OXZL.gif)
Example of a high-res, colour palette-restricted, pixel art animation.
Another main feature of pixel art is to be tied up to a restrictive colour palette. This is not always the case, but restricting design to a limited colour palette helps on giving the sprites a more nostalgic feeling.
So wrapping up, making pixel art nowadays has a lot to do with just drawing 2D art, with the added layer of working on non-interpolated images. This means any concept art could be transformed with a bit of work on a Pixel art software.
Also, pixel art has no size restrictions anymore, and it's just a way of making lines and shape more clear. This means nowadays we can draw really cool sprite art animations, like for example this animation:
![](https://i.gifer.com/OXZL.gif)
Example of a high-res, colour palette-restricted, pixel art animation.
Another main feature of pixel art is to be tied up to a restrictive colour palette. This is not always the case, but restricting design to a limited colour palette helps on giving the sprites a more nostalgic feeling.
1.2 Pre-rendered 3D models
And here it is, the method used extensively by Golden Sun GBA games!
So, first of all, what it's this exactly? Some people showed a bit of confusion around this concept and what it means, so lets have some examples.
![](http://www.nerdist.com/wp-content/uploads/2015/02/donkey-kong-country-pic.jpg)
Donkey Kong Country made 3D rendered sprites and backgrounds really popular
As you can see all of the assets shown in that screenshot are really 3D models. As the SNES didn't have any way of rendering 3D (with that fidelity, remember the Super FX chip), Rare just went full awesome and pre-rendered a lot of backgrounds and characters with keyframe animation based on 3D poses. That means once you have a 3D model developed you won't have to rig it or animate it, just deform the mesh for each needed keyframe an make a render with an alpha channel so then it can be integrated on a 2D engine.
![](https://gamefabrique.com/storage/screenshots/gba/golden-sun-03.png)
Golden Sun had the same pre-rendered 3D renders, just
with a low render resolution tied to GBA limitations.
The look that these games achieved is what also defines golden sun, as well as a lot of other games (Category on Wikipedia).
The way that Golden Sun gave life to this Pre-rendered 3D models is by having cartoony details on environment 3D models and doing really low fidelity on characters.
2. Team requirements regarding the preferred choice.
Here is where my main concern comes to play. A lot of us see Pre-rendered 3D as the best approach as it mimics the golden sun feel instantly. But it is really difficult to develop, in comparison with pixel art. Also, it involves a lot of expensive tools and software to make it easier.
We currently have one person that feels confident on doing spriting following this procedure, and that's ArticSin. We have SundaeS making some efforts but he stated multiple times that he doesn't really know a lot about 3D modeling.
The 3D modeling world is not the easiest to join, either. There are a lot of different tools, different methods for doing things, a lot of paywall barriers (as Blender is just madness to start 3d modelling...). So I'm concerned about if following this approach is going to not only slow down the project by a lot, but directly stop it. Without in-game assets we cannot make any progress in the actual game, and we cannot make public the builds as it right now uses copyrighted material.
So, in summary about Pre-rendered: To make this kind of sprites and background you need to know how to 3D model, how to light scenes, animating/mesh deformation, render techniques, consistent lighting and effects, consistent camera viewing angles, creation of render pipelines (that solves a lot of consistency problems stated before), texturing, normal mapping... etc.
Summary about pixel-art: You need to have basic how to draw. The less resolution we use, the more basic the knowledge has to be to make something usable. Also, it's really easy to start as there are tools that are really chip or straight-out free. In the other hand, copying sprites to learn and then applying some touches to it is a more feasible solution than doing the same with 3D modelling.
There was some discussion about using pixel art for some things and 3D models for other things. This approach could work, too. I feel we could do environments in pixel art and then npcs and characters in 3D.
Making a character to look like this:
![](http://img.photobucket.com/albums/v698/AtriusV/sprites/Aram.gif)
Involves creating a 3D model that looks like this:
![](http://img.photobucket.com/albums/v698/AtriusV/Random%20Screens/GSmaya.png)
So in a low-res pre-rendered scenario we could get away with doing some things on this procedure.
3. The problem of choosing pixel art and how it doesn't have to break the game feel.
So, as I stated earlier, doing assets the pre-rendered way just makes nostalgia instantly kick-in and make us feel like golden sun. That, tied to using similar animation keyframes, sound effects and so on, constitute a good foundation on how to make a game feel like golden sun.
But, due to the reasons explained above, what do you feel about shifting our in-game art to pixel-art based and try to mimic the original games feel in other ways. One of the first visual ways we could try is to make pixel art models seem viewed from a similar angle like the sprites on golden sun (nearly overhead view) and restricting our colour palette to what golden sun has. Golden sun has a palette debug mode to help us figure out how it uses palettes an what colours they contain. Link in here.
Anyway, I think discussing this right now has special relevance as we are just closing some character ideas and the engine is prepared to start using some sprites. So let me know how do you feel around this and how we should continue with our in-game art style.
On an end note, and if it's not apparent right now: I would do our game entirely by 2D sprites, tying the colour palette to ones similar to golden sun.
TL;DR: I personally don't thing going with pre-rendered 3D models is healthy for the project. I think we should shift to pixel art, but try to do that in a way it feels like Golden Sun, doing similar perspectives and colour palettes. Also, nearly everyone can start doing pixel art with a few hours of tinkering.