How can I get a good detail for my character texture?
I made some tests, and the results with 4k texture aren't crispy enough. I have seen a lot of renders in Artstation with 4k textures with a very high res result, Why my renders only achieve this kind of details with 8k textures?
The texel density is around 19px/cm.
The texel density is around 19px/cm.
Still, I would like to say that after testing, If I double the UDIMS to 10 instead of 5, and make each uv shell have a texel density of 40px/cm it looks much better. Is it optimal to do that?
Adding detail maps is not enough to make the texutre improve in terms of resolution or quality. The main texture is still to low.
The goal is to get the best quality that is possible to put the renders in my porfolio.
Criticism on topology and uvs is welcome.
thanks!
Replies
Could look for sample projects and check how they're setup.
Is it a game character?
If the goal is to create a game character, I would rather look at actual game assets for reference.
In both Unity Store and Unreal Marketplace, you should be able find sample projects. Can use those to learn from and reuse their shaders and other resources.
Keep in mind that high resolution textures alone don't make a good character. I would prioritise on finishing something.
Also I would make sure to optimize UV space to begin with by removing unseen faces (or have them reuse space), instancing repeating elements and mirroring parts where it is not apparent. And, as already mentioned, have varying density based on viewing distance and importance (hands touch face - should look consistent).
To find more on the topic, also try the forums wiki and search function. Good luck!
Right now the uvs are like this, with a uniform texel density of 20 px/cm and more padding in between uvs shells to avoid bleeding.
The UVs shown above in red/yellow/green/blue seem to have been made rather hastily, without much consideration for what needs to be straight or not and what needs to be kept continuous or separated. These would likely not be approved on a game production requiring sharp/clean detail ; and, parts of the model might have to be redone accordingly.
Usually a character in most games doest have a super even texel density, unless they are always seen at the same distance.
Meaning, in a fps game you will likely have lower arms and hands at significantly more texture resolution than the rest, in many cases those are entirely seperate models from the third person view.
In a fighting game where you look at the character from mostly the same distances. You might still have a higher texel density around the bust area for nicer closeup shots.
The example you showed is not from a game. Its realtime in the sense that it runs inside an engine. If you wanna go really highres, try to find models and textures from hellblade 2 as a really recent and very high fidelity project.
But in general define what kinda game your character should be from, the find references of said games.
Reuse all the texture space you can reuse, this can save you space on other parts which need to be unique. Traditionally this means, center line is unique, the further away from the centerline you go. The more likely it is, that UVs are stacked.
Next thing, straighten your UVs for better UV puzzling. You might be able to squeeze in more Texel density by better puzzling pieces together.
Don't zoom crazy in on the... knee?
Will you ever focus in the knee? No then don't inspect such things closer than you would see it ingame. What you seemingly are aiming for is more like vfx or movie style texturing. Usually very wasteful compared to actual realtime content.
The first step IMHO is to review the *model* and the way it's been constructed. Based on that one last screenshot there are a few red flags in terms of both design and execution : seemingly very tiny details, thin panels, and so on. Whether or not things will read well as a final game asset starts there.
This Viking character works because all the various elements have been very carefully designed ; and then further detailling and weathering has been carefully applied to them. It is a big amount of work, involving some expert juggling between basemesh modeling and highres sculpting/detailling.
Comparing this Viking asset to the glove screenshot you showed at the beginning of the thread is comparing apple to oranges, since the glove seems to merely be pulling its detailling from a one-click procedural leather material.
My previous comment about the base/ingame model still needing some work was assuming that you were going for a chunky, stylized look (which would rely very heavily on the model being very cleanly built). But if you are going for more of "high detailled cinematic" realistic look, then you'll also have to put in the time to create all this detailling one way or another - either through old school texture projection (Mari), or through sculpting, or through subdiv modeling. Possibilities are infinite, many of them involving some non-linear processes.
But you'll also have to be proactive about your own art direction, since the very cartoony proportions of your character aren't quite matching that of the Viking reference. If anything your project seems to be more in line with Clone Wars/Star Wars Rebels.
Art direction, model edits, further tweaks of the UVs ... This is a lot of work and a TON of inertia for someone still learning.
Therefore in light of all of the above my updated advice would be for you to focus *only* on one small part of the character (let's say the chestpiece+shoulder parts), and spend as much time as needed to bring that up to the standard of this Viking reference OR up to the standard of Clone Wars OR up to the standard of whatever is the most appropriate reference - with all the needed detail, weathering, and so on. Once that is done, you'll then be able to apply what you've learned (technically and artistically) to the full character.
And obviously, this design phase is best done with a solid concept piece guiding your every step. Improvising by just following the "tech" steps without a clear goal / strong guiding art reference is always of a waste of time and ressources.
But instead of a character from an animated movie, to prepare it to be optimal for video games. And I thought that the level of detail required could be achieved directly with textures in substance, without doing the detailing in zbrush.
This is actually the high and low poly versions of the model right now
and here I leave a more elaborate texturing test, so you can see a little bit of the texture work and the finish I would like to achieve.
Now that we know that you are *not* going for what people would usually think of when mentionning a "game character" (for instance : a hypothetical Fortnite version of Hiccup, which would involve geometry perfectly crafted for that purpose, and say two 4096 texture sets), but rather, a recreation of a cinematic version of it (using more of a UDIM-like approach without much concern for packing and perhaps using displacement maps rather than normalmaps) then I'd say you're more or less in a good place as it stands.
That's about as far as I can personally comment though, as I do not have much experience in that specific breed of CG. That said there are some great oldschool material about this out there : old Gnomon DVDs and Workshop videos about texturing characters for non-reatime film, which would translate nearly directly to a project like yours and might even open you up to workflows not necessarily involving baking everything from a highres source.
What's interesting about these (from what I remember from casually watching some of them) is that texturing in these cases wasn't necessarily 100% tied to locked/finished UVs, as the texture surfacing artist was sometime adjusting/redoing UVs on the fly as needed.