3D: Reverie pt. 1 (Thesis work)

So throughout this semester I've been working on my senior thesis project, currently titled Reverie. It's a puzzle game with homages to Myst and the works of Jean Giraud Moebius. The final products we're supposed to have for final crit, which was last Friday, are a trailer and screenshots, though making a navigable space is highly encouraged as well.
This has been quite the journey for me, as I chose a style that was quite out of my comfort zone. To try and emulate a very specific illustrative style, especially translating something form 2D, is something that I've wanted to do for a while. Here is some of the work that I made during the preproduction phase:
Concept Art:

 


 

 These were all done last semester, though, so I spent all of last semester translating this into 3D. I made a bunch of art direction documents to make sure that my environments remained consistent all around, especially in the Daydream environment in which the inkwork had to reflect Moebius' work very specifically.

These are some of the screenshots I had from the first blockout crit-- looking back, it's interesting to see how much they've changed.

 

For the Daydream. I'm using the Skull Advanced Cel Shader pack to achieve the lineart and colour, something that's proving to be a bit more finicky than I'd like-- the lineart is achieved by recognizing colour and hue difference, which would come to bite me in the butt later on when I was setting up the colour scheme in the Daydream.

Some of the first crit notes were about scale-- getting the desk in the laboratory shot to feel like it's a believable height was important, as was having the keyboard's keys the right size. In comparison to the abstraction of the Daydream, the lab needed to be super prosaic and realistic. Using these notes, I iterated on the next set of screenshots and began to sculpt the most important parts of the sets.

I like to think I'd made a significant amount of headway with regards to my key assets at the time. I needed to add a lot more smaller details, but that would simply come with working more on these assets.

At the time I was still figuring out the Daydream lighting; I knew it wouldn't be finished until I got the textures and materials sorted and then I could start making decisions about the colours and brightness. On the other hand, the lab was pretty easy to light: the main issue was what time of day and weather is best. I put up a vote to my friends as to which lighting scheme they preferred:

The results overwhelmingly leaned to the left, but Ryland pointed out that a middleground between the two would be ideal. So I compromised:

Finally, I figured it would be best for me to figure out how to do my textures as soon as I started sculpting them so I wasn't making any mistakes right from the get-go. Using the thumb, I baked it down and then used that normal map to generate a cavity map, which I then painted over to achieve this:

Which... wasn't so bad, but it needed work. One of my issues at the time was I didn't have access to Quixel Suite yet. nDo does the best cavity map bakes, which is really useful for me. I used Quixel Suite in most of my texturing, to be honest, as it gives me a lot more freedom than some other procedural texturing programmes. 

After a bit more modelling and texturing, I got my screenshots to look like this:

And then with this shot, I had arrived at a bit of a conundrum: which composition worked better? The first one shows off the fingers a lot better, but the second shot has a lot better movement to it. The first shot feels a lot more open, which is like Moebius' pieces, but the second shot leads the eye to the floating pavilion a lot better. It was a tough choice, but in the end I went with the second composition and I don't regret it.

And now some more of my sculpture progress!

The face console:

The lab console:

The landing circle (and how a really quick bake looks in-engine):

And then I took the pavilion and tried baking it and putting it in Engine. Despite some UV issues, I thought it was pretty cool.

I was really excited by how surprisingly effective these bakes were turning out, despite how quickly they were done. Once I had time to run it through Photoshop and clean up the lines, I might have a nice-looking building! One of the notes I was getting, though, was that the AO bakes from Quixel Suite were picking up some dark patches that really interrupted the linework, so I have to go in and get rid of those or they'll look like bad SSAO.

A bit more work on textures and I ended up with this:

 

I had to go in and really tone down the textured AO that came with the post processing volume-- while it looked really cool and gave everything a drawn feel in the shadows, it began to swim quite seriously when I moved a camera through the space. I was getting a lot of crit about how my lines on the baked assets, like the hands of the building and on the landing pad, were really sloppy so a lot of the polish pass was spent cleaning up my linework. I'm considering next time throwing it into Illustrator to see if I can get some really clean vector lines, but that'll be something to do next semester. For the Lab, I just kept making more assets to pack in the space and make it feel lived-in.