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
UV layout
triangle
All my UVs have been unfolded but I need some advice about the layout. In VFX we don't care too much but in gaming we want to optimize the textures to the max. For my character, hero model, how many textures and at which resolution would you suggest I use? I usually layout my UVs per object and grouped together nicely but there's a lot of texture waste. Again, in VFX, we don't care, we don't have the same restrictions as a game engine (To give you an idea, there was 200 UDIMs on the predator ship in the last movie).
So what do you suggest? How many textures? Should I just go automatic layout to optimize the layout as much as possible? (It's my first game character)
So what do you suggest? How many textures? Should I just go automatic layout to optimize the layout as much as possible? (It's my first game character)
Replies
- Making the UVs human-readable so that another artist can take care of the texturing (even if the texturing is driven by masks). Meaning, orienting the islands the right way up and grouping parts in a way that makes them as instantly identifiable as possible. Furthermore, orienting the various panels vertically and horizontally allows for maximum clarity of decals and graphics since they would then naturally follow the orientation of pixels.
- Taking into account the fact that the asset needs to behave well when the texture is downresed. Therefore having the appropriate amount of padding around the islands.
Here are some examples extracted from MGS4 (Raiden, Raven, Wolf), Overwatch (Tracer, WidowMaker, Vampire Reaper), Valorant, Zenless Zone Zero, Deadlock, and Arms. These span over 3 generations of hardware and a buch of art styles, showing that things really haven't changed at all over the last 10+ years or so besides the generational increases in texture resolution. In your specific case (futuristic spacesuit) you could probably find some good reference by looking up assets from Death Stranding, Horizon Zero Dawn, or any recent scifi shooter really.
While automatic packing can have its uses, it definitely takes a backseat in favor of the above factors in practice. If anything if two applicants had superficially similar portfolios but one used automatic packing on everything while the other shows care for manual UVs, the preference would go for the one demonstrating manual UVs as it tends to mean that the artist has more production experience under the belt.
(FYI, since this thread is not related to Substance but rather to modeling in general, "Tech talk" would have been the appropriate place and would get you more views. Or if you want to journal your progress on this piece and get help and advice along the way, it could go in "3D art showcase and critique").
This unwrap would not pass asset QA. Not just for reasons of texture waste but also because the parts themselves aren't quite cleanly unwrapped : you have symmetrical elements that are not symmetrical UV-wise (in other words : the "pelts" are not symmetrically laid following their center line). This may seem minor, but it goes back to the earlier point of making the asset rock solid for team work. If months down the line the asset is passed to a texture artist in charge of applying a graphic stripe to the chest they'd have to manually compensate for the wonkiness (especially if there is no projection tool available) instead of being able to apply the graphic symmetrically in PS. Hence a 5mins edit job turning into an hour.
Things like that do pile up very quickly, and it even starts to show on the final result when the texture is reduced for performance, whereas an asset with cleaner UVs would degrade more elegantly. To put things in perspective, if an external vendor was delivering UVs like these, they'd have to revise them fully and would risk not getting their contract renewed
Of course such things wouldn't matter if the UVs weren't to be shown in ones portfolio and not made for game use, ilke for a painted sculpt prototype with UVs done just for the sake of supporting textures, but for a game asset this isn't it (yet !).
There are more considerations too, like the way some strips and cylinder elements would benefit from being straightened (which can involve thinking about it very early, since it can affect the design itself), but I'm sure others will provide more specific feedback on that.
The left side is ideal, it's a lot easier to work with, identify and texture. While the right is a lot harder, doesn't fit the UV space as well and it will be harder to prevent a visible texture seam
Adding to the above comments, if this current body texture was to get scaled down to 1024*1024 (for instance if an hypothetical game was to be set to medium settings, or as the texture gradually loads and mips up during a cutscene), this is all that would be left to work with on the chest area in terms of pixel resolution :
While optimizing the UVs and straightening them out will definitely get you about 50% more pixel real estate, it will always remain limited and there isn't much one can do about it. That's why some gamers/laymen sometimes complain about "PS1 graphics" in modern games : it's not because the assets were authored at low resolution, it's more about said assets (probably originally authored with high res textures) ending up being reduced aggressively for performance ... hence ironically displayed at low spec in the end.
Like this door asset from the FF7 remake, which was probably automatically atlased together with other props late during production, and therefore getting its texture reduced way beyond the original intent :
This is of course an extreme case, and the artist bears none of the responsibility (as a matter of fact this door was well unwrapped !) - but I am bringing it up to highlight how fragile things can get in practice.
And to go back to the topic of straightening, here's another example from FF7. That's IMHO another case of unintended atlasing, but it's remarkable how poorly the big cylinders degrade because they were unwrapped at an angle instead of straight up :