My first "one week study" and my first serious attempt at a photoreal digital human in a realtime setting - or at all, really.
One Week Studies
First off, this project was the initial attempt at what will hopefully be many "one week studies", a challenge to myself, of sorts, that I started back in october 2019. The goal of these projects were:
About once a month, I pick a subject / goal to work towards learning more about in my spare time and within a week of that, publish the result.
Meaning I would have to stop working on it no matter what state it was in in the end.
I just have way too many unfinished projects laying around and figured this could be a good way to go about actually pushing myself towards presenting some of my work. Some of these will be from scratch, some of them might be finishing up a stagnant project and some might be a redo of old projects, but mostly oriented around learning new techniques and tech.
In fact, the whole idea of making this website spawned from the need to have somewhere to put these one week projects.
This project could be considered a mix of all three scenarios pictured above, as I had previously toyed with the idea of creating a british soldier character bust. However, my initial attempt done 3-4 years ago was never meant to be realtime, and barely made it to a renderable state. I thought I could salvage something from that project but ended up sculpting all of the head and clothing from scratch.
The sculpt progress throughout the week, starting with the model from 3 years ago, which didn't end up influencing the work much.
This new incarnation was going to shift the focus towards realtime, specifically utilizing Unreal Engine 4, and so I outlined my goals for the project as follows:
- Explore high quality realtime skin rendering
- Explore hair authoring options
- Explore high quality realtime hair rendering
- Explore high quality realtime cloth rendering
I started authoring hair in Zbrush, and while I like the tools there I found it too destructive for such an iterative project, and so I started looking elsewhere. I remember using Xgen in Maya at Uni, and figured I'd give it a go, but again found it too time consuming to update the hair with the constant model updates, with only the short week that I had.
So I gave Houdini a go. I've never used the hair tools in Houdini before, and although in a grooming sense they are much cruder than both Xgen and Zbrush, the level of proceduralism was enough to convince me. I was now able to update my model as I please with only minor tweaks required for Houdini to update all the hair and send it through a custom exporter I put together for generating the hair polystrips and different UV sets I wanted.
Overall, hair was probably the area I'd touched the least before and so I learned the most, but I'd also say it's what I'm the least happy about on the character, especially the hair on the scalp, eyelids and eyebrows.
The groom in Houdini on the left, and my custom hair exporter on the right
Most of my research on the subject of skin shading was done through Epics own documentation, specifically their section on Digital Humans. I found it immensely helpful, as getting the skin to look and feel right was very hard and I wasn't even close to happy about it until the sunday I was finishing the project. Substance Painter was used for the texturing but I ended up making some tileable textures using Substance Designer as well.
Here showing a closeup of specifically the skin progress over the week, and you can see how much I struggled getting the warm semi-transparent skin look until basically the last day. It also shows how especially the borders between eye and skin don't work that well and should've been given way more love.
Some engine screenshots from throughout the week.
Overall I'm happy with my results towards the end. There are lots of issues, and many things that I tried that didn't quite work out, like some of the scarring and the sweat streaks on his face. The eyes could have used a lot more attention, on top of all the time I already spent on them getting them to this stage.
The lighting ended up taking the backseat throughout the process and I probably could've put together something more exciting. I added the orange on around day 3 and it just sort of stuck with me.All my dreams of creating a more detailed environment for the bust started crumbling very early on so I had to settle for something simple.
The cloth rendering was also put on the backburner for most of the project, and although I did end up flooding it with "hairs" to see if I could create some more photoreal cloth effect it didn't really work. The fabric looks way too rough and bulky compared to the wool that it's supposed to be. The sub par cloth sculpting is also already limiting how good it can look, I think, but it isn't too jarring in the end result.
I'll round off with some shots that both display some of the things I like and the stuff that I don't.
I was surprised at how well the skin shading would hold up in the end. I didn't dream of being able to push shots this close before I got started. The face lacks a lot of proper skin details, but because of the characters perceived age it doesn't look too jarring. This shot also shows some of the lack of work around the eyes / eyelids.
Showing off some of the facial hair detail, blurring out the not-so-nice cloth in the background.
The rendering I'm quite happy with, and I'll probably do a post later about the techniques used to render these.