modelcorner
patreon
A more detailed look on Lanaya.
🕑 Added 2020-04-17 21:44:53 +0000 UTCHey guys!
Originally I wanted to make this post part of the release one, but I decided against, since both of them should be focused on its own thing. So in this post, I'm going to talk more in detail about the changes I did in. This model has many changes, maybe not too much in amount, but more in what it entails to make them.
Upscale model textures.
The model’s textures were pretty low res. The only choice I had was upscaling them using third party tools. The results were pretty good, and thankfully, since the outfit doesn’t really have any major detail, there wasn’t any detail lose with the smoothing of the upscale, so that’s a win situation for me.Change: Major increase in vertex count.
While this is not a big change since, as I’ve mentioned in previous post, I’ve artificially increase the vertex count on many model’s heads. This is so I can have more quality with the model, and have a better time with the flex building.
However, in this context, the increase was done on the figure. The difference is, even though I’ve made a post where I explained my line of thinking, all worth saying in here is I did a very selective Subdivision Surface manipulation: From the lower arms to the fingernails, from the calves to the toenails and the inner genitalia did not get any change in vertex count.
Doing that gave me the access to get the vertex count to more manageable numbers, which in order to be more specific, the objective is have less than 64k vertices.
In order to give some context, allow me to give you the following pointers:
- If you remove everything that isn’t from Daz, which is head, teeth, eyes and so on, Tifa’s vertex count from neck to toe, is 15,863 vertices, resulting in 40,207 triangles.
- In this model, if we do the same thing, is 31,863 vertices, resulting in 88,382 triangles.
Theoretical changes.
- By increasing the vertex count, by nature, it should result in better deformation. Keyword Should.
- The secondary effects of Subdivision Surface, means round areas should look more defined. This means in abs, chest and buttocks having their details increased, either looking more round or more “chiseled.”
- One of the negative effects that I aimed to prevent the most was messed up topology. By applying these changes and merging together two different poly count models, there was a huge chance to mess up the topology (“huge chance” meaning “it’s going to happen no matter what.) Another one of the reasons why I choose those particular areas, aside of not getting any single benefit from SubSurf, was because the Merge area was relatively low. The SubSurf arm has 80 vertices, while the regular arm has 40. While the location of all those vertices were a little moved around, the changes of topology between 80 (or 160) vertices is not that big of a deal. This is specially important because the lower arms are just a limb, they are not a main focus (like chest, buttocks, and in some cases, feet.)
This is also the case with the genitalia. It actually has the biggest vertex change, which would result, invariably, in messed up topology. That’s not an important matter though, because the genitalia posing is made by flexes, not by bones. This is why I left the genitalia out because, again, doesn’t get any single benefit from it.
Custom Normal Map for the torso.
This is not really a major change, but more like, a change I wish I could have the skill to do. I tried my best to have a normal map that would have let me say Yes, I like this. I wanted to do a custom normal map that made her abs look better, my skill however, go as far as understanding what a height map does, and not what a normal map does.
Edit: While I’m writing this, I thought about an idea in how to make a custom map. I will try it and see how it goes.
Results from the edit above.
So I had the idea of doing, again, the same idea I keep using that so far has always worked with me: create a regular white texture to paint it black and use that as a mask on external software. The issue I previously ran into when doing this, was that the edges were too sharply defined. You could see the roundness of the brush, which would make it look very weird.
I had a couple of ideas that I thought would help, and they worked. Or so I initially believed.
You see, before I decide to do any conversion, I always run a couple of tests in Blender. It is the reason why I’ve been able to understand how to do these things to begin with.
The results in Blender were really noticeable, and were also really of my liking.
When I did the conversion to VTF, the results were… at least for me, underwhelming.
Even though the I’m not even close to be happy with the results, I’m not particularly annoyed by it: it’s just how Source Engine works. Let’s keep in mind, Source Engine is old. VertexLitGeneric, as far as I know, is a pretty bad shader, but that’s how we have been rolling during all this time. It’s useless to compare a pretty fucking old game engine and expect it to behave nearly as good as the “Pixar inspired and PBR monster” Principled Shader of Blender. It would be like bitching that a lime doesn’t taste as sweet as a melon.
Something worth saying though, it is only a waste of time if you didn’t learn anything, and I in fact, did learn something. Now I have better understanding of the Height maps and how to edit them. I can’t tell if my results were good or not, but regardless of that, I like them, and that’s all that matters.
A step on the right direction.
I’m not afraid to say I deeply admire LordAardvark. I always share my ideas with him so, in my own words, ‘he can bring me back to my feet by telling me how stupid they are.’
I swear he’s just too kind to me he doesn’t dare to tell me that my ideas are garbage. When he released publicly his Daz v5 Preview, something caught my attention. If you guys are familiar with LordAardark’s works, you know he always share publicly both his ideas and his results (some in the form of text, some in the form of resources.) When he and Ordagon joined forces, then hell is let loose: you have LordAardvark’s impressive model work, and you have Ordagon’s impressively high Blender skills, only something good can happen. You guys know about how people give a lot of credit to them for their model’s butts. I do too, because Ordagon also shares something huge with LordAardvark, which is the openness to tell you how to do stuff: he used Subdivision Surface Modifier. I didn’t ask Ordagon to hold my hand in how to do this, he told me however, the modifiers he used and how he joined them. I was terribly impressed. And disappointed. Impressed because Ordagon’s skills are just too damn high (I’ve seen his practice works… jesus christ, he’s insanely good,) and disappointed because I was going to be unable to join it to my workflow because it’s horribly time consuming, and the chances of flex destruction are just out of this world high.
Fast Forward a few days later, LordAardvark tweets about the $maxverts command. As soon as I read that, I disconnected and went straight to test for vertex count. The export and compilation times are just oblivious, but that price is honestly, pretty damn affordable. I noticed the vertex count is just too high. I decide to think which parts should not be affected by the modifier. I reached a conclusion I have already shared with you.
Fast forward a few days later, the Daz v5 Preview shows some very interesting data: some parts are now affected by the same Subdivision Surface process, but more localized. I decided to ask LordAardvark about it, and ask him if he indeed added more SubD areas. The response was positive. I then, practically, forced him to read my conclusions from the SubD modifier (again, they are already shared with you guys) and it surprised me a lot: he reached the same conclusions, however, because he wasn’t aware of MaxVerts, he had to opt for the localized SubD. His conclusion was the same than mine. Honestly, it always makes me happy to reach the same conclusions than him while on my own, because that tells me that there’s at least something right in my line of thinking.
A plea.
I know I can’t tell you guys what to do, but this is the first time I will do it: please, use this model. I’ve suffered a lot building this model. This is the most painful model I’ve ever built, but that’s for the wrong reasons. I tried to get the textures because they came in the weird Source 2 format and I was unable to convert them. In my search to get this model working and get the textures I needed, installed tons of shit and a couple more, I ended up with a corrupted disk partition. I had to delete it, and with them, my own personal memories. When that happened, I actually wanted to quit everything. This was the first time that in my attempt to get a model to work, I end up with the most terrible loss yet. Yes, I lost my Steam account, yes I lost I don’t know how much money in games, but this time, I lost my memories. Those are invaluable. However, that same loss was the same thing that gave me strength to keep pushing. Yes, I lost my memories, but is that it? I’m going to just lose them like that? Fuck no. If I feel any respect towards myself, I’m going to finish this model. Yes, it is painful, but is worse to lose everything and give up. I rather lose everything but keep fighting. That’s in the past. The past won’t matter if I can’t look towards the future. Paraphrasing one of my favorite games of all time: Kill your past. You have already damned your future.
Anyway, this is getting awfully emotional and long, so let’s just end it with this note:
This was one of my Model Objectives. A Model Objective is “A model I’ve wanted to build for a long time.” And I finally did it.
