How detailed should models be?

[Calinou]

* (honestly, this looks way better than 99% of the models in FOSS games, yet Calinou thought it was justified to throw in some totally indefensible criticism?)

How is it indefensible to request decent texture/model resolution? We live in an era where graphics cards have gigabytes of VRAM, and stroage is bigger than ever.

The screenshot below shows Xonotic's Mega Erebus, a model made in early 2011:
The detail was pretty nice back then, but now it's a bit lacking, especially if you have Ultra HD or QHD screen resolution. Hence my call for slightly "future-proofing" models and textures.

This article may be useful: http://www.cgmasters.net/free-tutorials ... en-assets/

[charlie]

The zoom on that compared to leilei's photo is entirely different. Are you being deliberately ignorant or just not understanding the difference between what you posted and the image below?

[c_xong]

What really matters is how good the model looks in game, which is weakly correlated with the number of triangles, since you can cover up defects with surface effects or magic like runtime tessellation. It's really hard to make an accurate judgement when someone has basically posted a disembodied work-in-progress model.

[leilei]

We live in an era where graphics cards have gigabytes of VRAM, and stroage is bigger than ever.

We also live in an era where there is the Raspberry Pi, smartphones with relatively poor 3D, integrated 3D, poor BSD drivers, bandwidth usage caps, and idle "ideas persons" that think they know better than those who have been at this for over a decade who learned things the hard way.

Here's some things that will offend you:

- Minimum ideal target system spec is a Pentium II 300MHz system with a 3dfx Voodoo2 card.
- A player gets 2500 triangles maximum and LODs are required.
- A player gets a single 512x512x24 texture maximum
- There is no normal mapping or real-time lighting.
- Nobody will glow a solid primary color for most of their character.
- There are no PNG files and never will be.
- I'd like the game to not run like crap. Texture switching on modern low-end hardware is a big deal.
- Target size is 100mb. I am not interested in a bloated game.
- Time isn't free. I only get a few hours a day to work on this project. Alone. Sticking to diffuse-only meshes helps this.

The targets are set in stone. This is not up for debate. I'm not changing my targets and engines, throwing away years of hard alone work because you want a smoothed glowstick arena game (as if we don't have enough of those boring things representing the majority of FLOSS 3d action games).

[Julius]

- Time isn't free. I only get a few hours a day to work on this project. Alone. Sticking to diffuse-only meshes helps this.

This! Although to be fair to Calinou (who is not just an "idle idea person" by the way ) adding a few extra polygons to smooth out edges really does not take much time at all and even on low end hardware has little impact these days (hardly any systems, even the ones mentioned by Leilei, are geometry transformation limited in a typical game... but optimizing for texture fillrate is indeed vital as far as I know).

But overall this is a stylistic choice and I agree that the new OA3 models Leilei is working on look fine for what the aim is.

[Duion]

Don't use more polygons than you need, but also don't overly save polygons, except if you target mobile devices or similar, but otherwise you can have millions of polygons nowadays.

[Vandar]

I think, in the beginning of a project you should aim at the top level of the technology (as many poygons as good current cards can do) since until your project is finished, the top technology from today will be the average.

The tricky questions are to estimate how long your project will take and how far technology will advance in that timespan. But you shouldn't use todays technology as a reference but the one from the day when the project is done.

[leilei]

No, that's not a good idea. Even if you take that idea into the context of 2005, or even 2001, you'll never get anything done.

It's always easy for one to say "use normal maps up the poly's tighten the graphics ehre great easy way to make character it's Make Human!" without regarding the lack of vertex shaded skinning, lighting transforms, memory allocation issues, higher human labor requirements... there's a reason why some of the best looking FOSS games out there are ex-commercial games once done by professionals (i.e. 0 A.D., Nexuiz)

I have to have a game that hits the framerates into the 300-500fps on common hardware, and i'm already playing with fire by having a 2500 triangle budget on players.


also... art direction > technology

[Andrettin]

I think, in the beginning of a project you should aim at the top level of the technology (as many poygons as good current cards can do) since until your project is finished, the top technology from today will be the average.

For an AAA title, yes. For an open-source or indie game, that would make finishing the game probably impossible given the cost of producing those assets (or the time required to do them yourself).

[Arthur]

...there's a reason why some of the best looking FOSS games out there are ex-commercial games once done by professionals (i.e. 0 A.D., Nexuiz)

Nitpicking a little, but 0 A.D was never commercial, it only began as a mod for a proprietary game and then became standalone. As far as I know, very few of those involved have worked professionally, but the result certainly has become of such quality art-wise.
Also, they do use normal maps.

art direction > technology

So very true.