Week 4: A kind of 'Game Dev Post Mortem'?
So this past week has been a flurry of emotions and in a way, I'm sad to have finished the group project but incredibly happy it's over and done with. Being with an awesome group of people really helped lighten spirits during the dark days of light maps. Gah.
For this project, I had to work within a
group of 5 people in order to create a replica of a chosen film scene. Other
than the fact the textures had to be 1024 x 1024 or below, there was no texture
budget (awww yiss) and our tri limit was 20,000. This is also the second project I’ll be
using PBR for so this is a chance to really get to grips with it.
Each member of the team started off by
compiling various moodboards of stills in order to present to the ‘Dragon’s
Den’ board. These contained stills from films we all liked ranging from Fight Club to Sweeney Todd. After about 4 days of deciding and having previous ideas shot down,
we finally settled on Moonrise Kingdom. The critical studies team seemed to
like this idea – although they did say the textures may be a pain but I’ll get
to that part later. After making the final decision, we really
knuckled down to do concepts, more in-depth moodboards, colour pallettes and
whiteboxes produced in 3DS Max. The research also included a plan of who's
going to be modelling what, personal deadlines and so on.
 |
| Moonrise Kingdom board |
After the completion of the research, we then
all moved onto the modeling stage. We’d been shown how to use the perspective
tool in 3DS Max which was extremely useful for getting everything in the exact place, if
not a little confusing at first. This was also a complete and utter nightmare at times
due to the scene being shot with either a wide angle/telephoto lens or *possible* forced perspective. Could be a mixture of both. Who knows?
I was given the following assets to model:
- The two tables by the stairs
- The two lamps on said tables
- The board game and associated assets
- Finally, the Window
They were all really fun to model and I
stuck within the tri budget I’d been set which was a relief! The only object
that went slightly over was the board game. Due to modeling almost every day
over the summer, the speed in which I model has increased so much so I managed
to get all of these assets modeled within a day. Progress!
 |
| Excluding the window, here's the assets I modelled |
Now onto the best part of a project…unwrapping!
Again, it also only took me about a day and
a half to get everything unwrapped, packed and sort out my light maps.
Everything was going swimmingly until I had to make light maps. For some reason,
there’s an odd bug in Max that will delete the original UV channel as it sees
fit if you don’t keep collapsing the stack after moving channels. We did all
manage to find a fix for this in the end though. But it took a LOT of
collapsing. There's so many bugs I've encountered in 2014 that it's starting to become rather amusing.
I was actually quite worried about
texturing to start off with as I’d only done one other project using PBR and
although that was simple enough, something was bound to go wrong here. I wanted
to set myself a challenge by hand painting all of the textures myself – which
I’ve really grown to enjoy. I also find this much easier to work with when
using PBR as it’s extremely difficult to try and remove all possible light
information from an image.
When making all of the textures for the
table and the window, I had a base colour and then created a new layer and used
the ‘fibers’ filter in Photoshop to achieve a grain effect. Although with this,
it was only straight grain, so I used liquify to make the larger ‘circular’
areas of the grain. I mentioned earlier that I’d baked my normals but there
were various problems occurring along the way, so I scrapped the original plan.
Instead of baking, I made a height map in which I managed to create both
roughness and normal maps. Baking is way more fun. :(
Another thing I've discovered throughout the texturing stage is that Marmoset is one huge lie. I used it to test a couple of textures on my models, and it looked totally different in engine! So instead of wasting my time and having to adjust materials twice, I just whacked everything straight into UE4 and started building the scene from here. Why Marmoset whyyyyyyy!
 |
| Lamp tested in Marmoset (LIES) |
Once I'd imported all of my own assets, I offered to put everyone else's in too so we all had one scene containing everything. Although once i'd imported the lamp, I soon realised that the alphas weren't working. Anything else want to go wrong? I tried many different methods and re-exported everything but it didn't fix the issue, gahhh. In the end, I gave up and made a glass material in UE4 which looked awesome! Luckily, we'd all sorted out our own light maps so we knew that there wasn't going to be any issues here. However, once I imported the structure and changed the view to light map density in UE4, the whole scene turned red. Perfect.
We tried changing the resolution of the light map, and managed to get the walls green but everything else was still red. Even better. The only way we could resolve this was to import the staircase, the walls, the floor and the ceiling all as separate objects. Ahhhhh it just keeps getting better. However, this did fix the problem and everything finally turned green again!
I realise how much this post has dragged on...so I'll stop talking now. Here's some shots of our scene!
As you can see we've still got a bit of an issue with shadows on ceiling and the floor, but it looked pretty damn sweet until we rebuilt the lighting, but it's gotta be done! Overall I've had an incredible time on this project and I'm really starting to love group projects too. I miss it already. :(
 |
| Allllllll the greeeeeeeeeen |
Anyway, I don't want to bore you with light maps. The actual lighting is much more interesting and pretty. We all had a play around with lighting in the group to see how it affected the assets and so on. However, when it came to the final lighting pass, I did this due to other members of the group working on different things to add to the scene. I think this was probably my favourite part of the project and being able to do a job as important as this was absolutely terrifying as I didn't want to let the group down, but I did it. Eventually. I had a lot of feedback from the team, which really helped and led to an estimate of 8-10 iterations of lighting overall. Woo. To start off with, I imported the base mesh without any textures to get a feel for how the lighting works in UE4. Luckily I practiced on UDK over the summer and the difference wasn't too horrific to say the least.
 |
| Final Render |
The only other thing other than the film project this week was 'Cool Shit Wednesdays' which was a few hours of watching our lecturers and other students scream playing Alien: Isolation. Although we still had the presentation to make, it gave us all a chance to chill out for a couple of hours and totally forget that we had a huge deadline the next morning...
0 comments:
Post a Comment