The BRAWL² Tournament Challenge has been announced!
It starts May 12, and ends Sept 12. Let's see what you got!
https://polycount.com/discussion/237047/the-brawl²-tournament
It starts May 12, and ends Sept 12. Let's see what you got!
https://polycount.com/discussion/237047/the-brawl²-tournament
Mesh resolution
triangle
on a AAA game, how much of the details are modeled VS normal map? Like on the helmet and the armour.
Replies
- But also keep in mind that while this breakdown does show somewhat of a "bread and butter" approach to game assets as commonly seen in current highly detailed productions ... the amount of work involved for a mere headset prop can also be viewed as the reason why game budgets are becoming absurdly expensive, as you'll notice that building a real-time game version of the item is *more* complex than building the CAD model done for the production of the actual real life item. Yet there are other ways to create game assets - like not baking anything, or leveraging completely different modeling and texturing techniques. You'll find these approaches particularly in visually stylized eastern games and in some games produced under strict technical constraints (for instance game originally made for mobile of for the ps3/ps4/switch and targeting high framerates).
- FWIW here's another breakdown dating back a decade. You'll notice that the main difference is that the ingame model is more coarse, but besides that the steps involved are pretty much the same.
At the end of the day, games performing very well are the ones leveraging clever optimizations all the way down to the way individual assets are authored.
In this model, the pink part is soft but all the other ones are hard surfaces. Should the soft surface go all the way under or I just keep the polys I need so I don't make the model heavy for nothing?
Being wasteful is of course never a good idea but that's not the deciding factor for this kind of stuff and it has not been the case for about two decades (in the sense that keeping or deleting a few hundred tris keeps a model within the same order of magnitude density-wise anyways, hence would likely not blow up a tech budget). What matters is the actual context the model is used in game, the way it behaves in motion (with no way to hide things per-shot or in post
For instance had Epic decided to have the hard shell panels float on top of a soft bodysuit underneath, this would mean a significant chance for some unintend clipping and interpenetration creeping in under extreme poses especially with third-party animations from the marketplace ; and their LOD generation might also have been compromized. Yet Valve had some different constraints driving their decision of shoving parts into each other, the main one being probably ease of editing since their setup allows for later adjustements to the ingame model without any rebaking, whereas any edit done to the Epic UE5 mannequin would naturally require a complete rebake with new UVs.
This is heavily context dependent - but it's a robust rule
there are of course other reasons to have geometry.
one thing I will add is that shoving the arm in to the glove (as per the example above) not only risks unwanted interpenetration etc. it is likely less efficient to render than the watertight alternative (cos overdraw)
and yes - weight painting is usually enough - you might need to add some supporting geometry for awkward parts
For a lowpoly I'd argue that the mesh density/distribution is odd
Clean and simpler parts being denser than other parts right next to them
For a lowpoly i would probably mesh everything together, kill all the chamfers and let those be carried by normalmaps unless you intent to go really close. And even then, a normalmap likely would carry the chamfers
To give you some sort of metric..
The ideal from a rendering performance perspective is that no triangle is smaller than a 2x2 pixel area on screen so given an expected view distance/resolution you have a measure for how dense your densest geometry should be.
Rather than measure it, I zoom until the object is roughly the right size on screen and look at the wireframe - if I can't see gaps in an area(eg bevels) then I probably have too much geometry for that view distance.
By that I am not saying that this is a bad target in and of itself (as it will certainly make for a fine portfolio piece) ; just saying that right now, you would be better off focusing on the shoe alone for instance, as a test case for the whole pipeline (including : retpopo, proper handling of triangulation, baking, texturing, weighting, and realtime presentation).
Otherwise you'd end up spending days or weeks on the highres of the full outfit ... only to realize that there is still a whole bunch of unknowns waiting for you after that. Whereas spending a week or so fully focused on a small part from start to finish will force you to run into all the specifics of the pipeline early on, and you'll be able to apply that to the full character later knowing full well what's ahead.
Is it possible to transfer a Maya (or Blender) rig directly into Unreal using FBX and keep the weighting?
You'll often see only senior openings, or not even senior openings yet all asking for 3+ years of experience in that position or N AAA games delivered. This speaks about an excess of candidates for few openings, and in those cases they tend to pass over very experienced people from tangential areas, or from the wrong country or whatever in favour of the candidates with a similar or equally good portfolio that checks all other requirements on paper, because they'll get plenty applications from them too.
That's kindof my point : by tackling a tiny portion first and taking it all the way down to the finish line (instead of doing a full character), you'll force yourself to run into this kind of question earlier than later and you'll save yourself hours/days of wasted time. For instance, to go back to your question, you'll see that UE4 and UE5 come with example projects featuring standard skeletal male and female characters that can be exported straight out of the engine to use as reference, that you likely want to base the proportions of your character on (especially since these default UE5 mannequin characters happen to match the proportions of Epic's Metahumans).
If anything that's the unique thing about game art : a big part of making viable assets comes from being aware of the specifics of a given pipeline and/or engine and working with that. For UE, that means being hands-on with the default mannequins. That's why modders and UGC creators (very familiar with what it takes to put assets into a game, either through third party or official tools) make for good hires for a team. Of course on the daily a modeler would likely not be in charge of implementing assets in engine, but it's good knowledge to have.
As a matter of fact once you're done with this spacesuit I'd recommend tackling some Warframe and Dota2 UGC content. Warframe UGC would mean making a head replacement using a completely different toolset for implementation, while Dota2 would involve crafting a modular outfit around the base body of an existing character, under a very strict budget and yet another toolset.
Also FYI the UE5 editor is significantly heavier to run and less responsive than UE4. Some users don't mind it, but I am mentioning it because it can get in the way in some cases.
(And, thanks ! It's due for quite a bit of reorganization soon).