Recovery, 3d learning, uncertainty.

Recovery so far has been a bit iffy. My health issues have improved, but things are not as normal as I'd hoped they might be. As a result I've found it a bit difficult to get my bearings. There's the constant feeling that I can't relax, and it's difficult to bring myself to focus on work. It's also not often that I feel as though I can commit to streaming (though I will still probably try soon). Inevitably, I end up working on things I just feel like working on that day. While part of the problem is certainly still physical, I need to find some way to get myself out of the rut I'm in.

These days, when I don't feel like working I've gravitated a lot towards learning 3d. I haven't posted in a while, so I figured I may as well go over some of the things I've worked on over the past year, any time I've worked on it.

3d stuff

Thus far, my goal when working on 3d has been to figure out what I would do if I were ever to make a 3d game. Learning the workflow, trying styles, and deciding what kind of projects I would work on. Because I would likely be working by myself, a 3d project would have to be viable for one person.

For many of my early tests, I attempted to stay low poly and minimize texture work. Though these rarely remain low poly and inevitably get more detailed.

Figured out how to do weapon trails in Blender, so that attacks can be animated with a preview. One of the reasons I prefer stylized is that the bar for animations is different. Things can be fairly snappy. Run animations clearly need more than 8 frames though. I tried a lazy one here, and it was too basic.

One of the things I would have to decide on with this sort of style is if the face is drawn, or modeled. Modeling the face is more time consuming, however it does allow you to use shape keys to animate it. With drawn textures, you would need to use some UV tricks.

Tried modeling some low poly armor in disconnected parts. Certainly got me thinking about Dark Souls.

I had recently played Demon Souls, which made me want to try making a mockup of an area. So I started thinking about how I would go about more "realistic" graphics.

I wasn't sure how I would be creating textures, so I looked into some different options. I began using software called Quixel, which allows you to generate textures using existing images, or from scratch.

I decided to try a more "realistic" looking character. 

I generated a texture for hair using Blender's particle hair system.

Using this texture on planes, you can make fairly convincing hair. Using a high enough resolution texture also probably matters a lot for alpha clipping transparency.

Rather than model the back of the head, I tried a method where I leave it blank, and shape the hair first. I think this method is the easiest way of creating hair, for me anyway.

While creating this model, I also realized that there is a fundamental difference between realism and stylized. The stylized method is usually to make things "the color that they are", and you use shadows to add some color. It looks like this,

Whereas if you make the color very dark, and use far higher light values, it looks like this. In general this is more realistic looking because you can use high light values.

Tried a common method wherein textures are automatically applied based on angle. It seems many games are textured like this as a base. It's a bit nuts what you can get away with using only 1-2 textures.

I realized that you can use Blender to generate height maps, using models. This can be used to generate height/displacement maps in Quixel. 

Because Quixel allows you to generate maps from images, I also began taking photos to use as resources. These can be used individually or be combined.

Was amused that using textures without the normal maps kinda looks like a 2000s PC game

I did a lot of tests using different levels of exposure. Seems like finding the right exposure and light settings is as important as many other aspects.

When doing these tests, part of the goal, (contrary to making an amazing looking scene), is to try and find what the minimum amount of work is to make a lazy area look "complete". Referencing AAA games, there are honestly a lot of games with sloppy texture/environment work. Not to say that you should make a game look bad, but it is at least a little reassuring to know that games you enjoyed a lot were far from perfect, and used a lot of the tricks that are available to you. 

Something I noticed when referencing AAA games, is that a common methodology is to model the environment as a trail, which is walled in. This allows you have something like a forest without having a million trees and bushes blocking your view. You put all the props on the sides. This design philosophy is also present in things like Dark Souls, despite those games having a lot of exploration.

I started modeling an enemy with hard modeling (vert by vert). It seemed like it wasn't very intuitive to design detailed enemies while also modeling it and worrying about topology. There's also the issue of normal maps, since you need to generate the normal map from something.

So I looked into the sculpting workflow. Essentially you do an initial sculpt, auto retopo it into a low poly mesh, shrinkwrap it to the original high poly sculpt, and it leaves you with a low poly model you can bake the high poly sculpt to, retaining the sculpt details. 

After sculpting, I can certainly see why monsters tend to look the way they do in modern games. The sculpting work flow makes certain kinds of shapes very easy to produce. It's also easy to make a mess of shapes and simply see where it leads you. There are probably a lot of bad habits to fall into here.

Textures like this in the skin can be made with a brush and randomized rotation. This one is actually a brick texture being used as a brush. It seems like in general, one of the skills in sculpting is knowing when to use large details, and when to use small details, since it's not like the enemy will always be right next to the player, and they should read well at a distance.

And that's the most recent model I've worked on. Honestly, I've been kind of surprised how easy the workflow is. We're lucky to have some stupidly useful tools, like ones that generate maps from a single image, or auto retopo software etc. At this point, I'm not really sure what sort of style I would go for in an H game. Originally I figured there's no way I'd do modern/realism, but the workflow is actually pretty simple. Though ultimately it is still slower.

I'm also not sure when I will be working on 3d. If I eventually do bring myself to stream my work, perhaps I should set aside one day a weak for 3d, and stream the learning process too. Dunno yet.

Anyhow..that's all for now. Hopefully I will bring myself to get back to work, and stream it at some point. I need to do something to break up the state of things, though.