A Sidetrack Into Pre-Rendered Games

I loved the early games of the 90s and 00s which relied heavily on pre-rendering 3D models into high quality 2D sprites. Last week wrote a bit about 2D game interiors and one that stuck in mind was Disco Elysium. I had hop back into that game to appreciate all of the pre-rendered, isometric, hand painted beauty of what those backgrounds gave us. As I did more research, this idea of making something pre-rendered popped into head… hey what if we gave that a try?

Isometric projection was originally a clever workaround for hardware that couldn’t handle 3D. They chose to use isometric to prevent perspective distortions which allowed developers to build massive worlds in a sort of fake 2D perspective. So many good games like Diablo, AOE, Roller Coaster Tycoon to name a few. A lot of games have transitioned into fully 3D because of hardware advancements, but achieving that level of detail at the time in real-time was something both beautiful and magical. Even in modern indie dev, I can see a resurgence in pre-rendered with games like POE, Hades, and of course the earlier mention of Disco Elysium. Nowadays devs are combining both 2D pre-rendered, and 3D assets together, using the best of both worlds to get a unique aesthetic we haven’t seen before.

Trip Harrison has a great in-depth writeup on The EnduringBeauty of Isometric Artwork

And so… without much experience doing anything like this, I ventured off into Blender to model a small scene with a house and some of our flat foliage on 2d cards. Cozy right? This clay render would serve as the basis for the pre-rendered isometric camera angle…

I probably could have stopped here and brought this into Unity to complete the test, but wanted to take another quick detour into Midjourney…

(I know, I know… AI for art… blasphemy!! Unsubscribe now! As this is only a proof of concept prototype and none of this will be used for the final art or final game, I don’t see any harm done here… Resubscribe please. Side note, I do think this one way AI image generation could be used fairly. Using it for conceptual iteration, or reference, or to just spark some ideas has its place… BUT not for final art. Also if I had AI do all the work for me, what’s the fun in making art in the first place right!? )

I diverge, continuing on… Midjourney Retexture was used on the 3D render from Blender to create the image below. It’s not exactly what we envision for the look and style of the final game, but it has some of the right feels. So after lots and lots, and lots, I mean lots of random rolls and re-rolls, we get this…

With all of the AI fudgyness and melty globs and gloops, it’s somewhat descent (again this is where you can see AI failing to as final art). Ideally I’d paint this pre-rendered image, or texture the model in 3D just how we’d want it, but a MJ generation will have to do for now. Now… into Unity we go and a proof of concept was born. Add an invisible shadow catcher plane and you’re good to go right… not so fast…

The character was walking around and it works! I was surprised at how convincing the results were but it still wasn’t quite right as you can see from the video above. Since I already had a 3D block out model of the house, I brought that in as another invisible mesh to catch the shadow on the walls and also use as collision to block the player…

It worked!… kinda. There is double shadow (on the floor and on the house wall) and still no ability to occlude the character behind the building, no way to interact with the plants, no way to go inside. This was tougher than I initially imagined. Once I turn on the invisible mesh, all you have is a floating plane, perfectly aligned to the isometric camera to make you think this is a high resolution 3D scene. It’s still all a FAKE!

The initial setup was pretty straight forward, but after a couple of days of troubleshooting, debugging and trying to sort things out, I realized this was going to be more of a hurdle for the game. It probably wouldn’t save us more time and in the end would probably cause us more problems and challenges.

As an experiment, it was a successful prototype and proof of concept with a lot of technical hurdles to overcome. At least now, I know we’ll probably won’t be going this method even though it opens the potential for some really unique art styles and beautiful art without any performance limitations.

I have a new found respect for the games who ended up re-rendering their game, and a bigger respect for the devs who spent time painting, integrating and making it work!