https://80.lv/articles/overwatch-hq-creating-a-game-level-for-portfolio/
"Overwatch HQ" - Long-Term Environment Working Thread [UE4]
polycounter lvl 10
Latest Update
I'm designing and building the Overwatch Headquarters which is located in the Swiss Alps. Check out the 2nd post for more explanation.
OWHQ Jan2018 Progress Walkthrough:
https://www.youtube.com/watch?v=D8uwYfSW9tE
OWHQ has also been featured on 80 Level!
All images are still works-in-progress!
I'm designing and building the Overwatch Headquarters which is located in the Swiss Alps. Check out the 2nd post for more explanation.
OWHQ Jan2018 Progress Walkthrough:
https://www.youtube.com/watch?v=D8uwYfSW9tEOWHQ has also been featured on 80 Level!
All images are still works-in-progress!
Replies
Good luck!
@CptAlbatross And you too! Thank you so much. I've learned a ton since embarking on this project, both in the backend thought process as well as in the practical sense.
And here you've hit on one thing where I feel like I've really kind of dropped the ball in this project, haha! My Maya files are not as clean as they need to be!
My blockouts and prototypes are a mess, but I quarantined those maya files as well as the fbx assets and scene files to their own (as per the org chart from page 1)
From there, I went to the vertical slice, so I made a "MorrisonsOffice-master.ma" file, and I even kept that one relatively clean. Everything in environment art needs to be correctly sized relative to everything else, so I like to work in as few files as possible, but what that means is that the files internally need to be meticulously cleaned. (Maya dumps so much garbage nodes and empty groups and such). When it came time to model geometry for a high res normal map or duplicate a bunch of iterations of a prop from start to final, I'd usually just grab all the meshes when I was done, group them, name the group, and then hide the group. Then I can grab a duplicate of the final asset, name it proper, and place it in the scene to work with it. That way I can scroll through my outliner and see all my objects, as well as all the little hidden projects that happen around in the void.
I can go back into those BAKE groups and grab high res geometry to use in future bakes or whatever. It worked really well.
Unfortunately, when I shifted off of vertical slice and into production, I made a "Hallway-master.ma" file which became my main working file for a long time. I used it to model everything in the interior as well as blockin the skydome and get all my sizes correct, and the most organization I did was throwing certain collections of assets (ceilings, some walls, skydome geometry, etc) into layers so I can turn them off for visibilty. This file got huge, and I got lazy and stopped naming things. Then, one day, something happened while saving that file and it was suddenly rendered corrupt and I freaked out! I was texting a tech artist friend of mine who knew how to salvage maya files. What redeemed me was that I had been saving as Maya Ascii from the start as opposed to Maya Binary. If an MB gets corrupt, it's hosed, but if an MA goes wrong, you can open it with notepad and hopefully find what's wrong and fix it.
I was able to find where the corruption began and delete is and salvage most of my file, but I lost all of the materials that were in the file. Now, this isn't too big a deal since in the UE workflow all material application and such happens in-engine, and all of my UVs were still intact, but it was super annoying to come into my main master file and have everything be neon green missing textures and such. That file has never run very well ever since and I definitely get more crashes than I did before. Dumb.
Now that I've started rendering out the exterior, I've created an "Exterior-master.ma" file and I'm keeping that one clean like I did in the beginning. When I move on to the skydome, I'll be creating a Skydome-master as well. It really really pays to keep thing clean in there, even if it takes some work away from actually producing assets. I think for me and the way I work, I really like to just work in 1 file as opposed to putting everything in their own files and referencing them in, but whatever is cleanest and works for you is good, as long as you're making those decisions at the start of the project and they're decisions that become Policy that you will keep to and work with.
In short,
Do this:
NOT THIS:
In other news, I'm grateful for OWHQ to be featured on 80 Level today!
https://80.lv/articles/overwatch-hq-creating-a-game-level-for-portfolio/
Excuse the mega noob question ha, but how do you export your assets? Do you export the whole office for example as one mesh? Or export all its pieces and then reassemble in UE4, just copying the coords from maya? Or something else. Being pretty knew to game dev, I've always struggled with the best way to deal with handling multiple meshes.
Cant wait to see more.
You definitely don't want to export a whole room or level as one because that won't play well with how Unreal assigns material groups to an asset. The more materials you have on a single object, the more expensive it is, so you really need to be mindful of how many separate materials you have on the object you're exporting. As such, it's probably cheaper to cut an object into a couple pieces with a few materials on each and export them in-place so they're aligned when moved to the origin than it is to combine it into 1 object but have a dozen materials on it, if that makes sense?
@Nuna Thank you! And you are absolutely correct. If this were to be a real level, frankly speaking, it would need a whole new layout taking into account all of what you've laid out. As it is, though, I've made the call that since this is an art project, it's more made to look like it's a real level, but I also wanted to adhere to some of the layout in the Uprising comic, so I've tried to split the difference. I think my next project will be more of an actual, real layout level, though. The closer I get to something being actually playable the better I feel.
You'll also notice a lot of bespoke, story hook props on the task list, and I'm thinking about streaming myself making some of these if people would be interested. Probably Tuesday and Thursday nights or the like. My channel is here if you'd like to follow it and get notified when the streams go up
https://www.twitch.tv/mutatedjellyfish
I know everyone and their dog streams, so these might just end up being me silently modeling by myself, but y'all can always watch the saved videos later, and if people do show up I'll be more than happy to talk about whatever people want to talk about in regards to art and art making, including my process, or staying sane, or whatever!
I can rest my eyes on these scenes.
After fiddling with various material styles, vertex normals, and shadow options, the grass has landed there for now. I'm happy with how it looks in sun and shadow now, even though you can see the cards pretty well. The low angle of the sun casts really long shadows which doesn't help, haha. Let me know what y'all think! Don't worry, the trees are all placeholder, so they'll change. They're a big hot mess for now.
@Archsider Haha, thank you, goodness! I'll gratefully take the compliment even if I don't entirely agree.
Aaaand done. I promise I wont make this thread into a place to post about streaming. Once I get a set schedule down, I'll just post it in the OP and be done with it. This was my first time streaming anything ever, though, so thanks for checking it out of you did! I didn't get much work done, but I sure talked a lot.
Another thing that has been stealing some of my time has been prepping for GDC! Given my employment situation, I've decided that this year would be a great time to go to GDC for my first time. I've always wanted to go, but now it makes sense, so if anyone is headed there as well and would like to talk, PM me or get in touch via some other means.
Moving forward, off-stream, I'll be continuing to dress out the ground plane. The terrace has much more open, navigable space than the interior, which means I'll have to find my interest and variation along the ground as much as in the surrounding elements, so I'm thinking about things like a few dirt/grime decals, dropped pine needles, some light debris like papers, etc. From there, I desperately need to address the trees and railings and then the more bespoke props like the flags, exhibit pillars, and that infamous statue!
As a reminder, here's the sketch of the Alpine Terrace:
And for now, if anyone is interested, my stream will be up every Wednesday and Thursday night, starting at 9pm MST.
Edit: Oh geez, I only just now noticed that putting a raw link to a twitch stream in your signature means an embedded stream window at the bottom of every one of your posts. I'm sorry, guys, that's super obnoxious.
First off, streams! I stream Wednesday and Thursday nights and I've dedicated the Torbjorn makeshift workspace out on the terrace to be my stream work and I'm starting to build some momentum there:
Clockwise from left: Computer work station desk, computer server stacks, futuristic digital drafting table, workbench, toolbox, wheelie cart, storage boxes. The idea is Torb has dragged all of this equipment out onto the terrace to set up a makeshift workshop. I've also begun blocking in lots of the simpler-styled props and population that we'll need to really make this area feel used and lived-in. Not pictured yet is a man-shaped punching dummy that's been drafted to use as a mount for Mercy's damaged wingpack, which is what Torb has been working on repairing out here. The chat on stream helped me come up with a prop list for the space, and most recently, I've been setting up a simple texture atlas/material instance system in order to simplify and streamline the props and their production so hopefully we'll be able to make this feel pretty rich. Also present are my newest stencil objects: spare papers.
The stream is challenging for me in a lot of ways, but where I may lose speed, for example, I more than make up for it via the chat braintrust. Viewers (you know who you are!) have called me out on silly things I do wrong and taught me all kinds of new things, especially about UE4 and workflow. I've even had moments where I've stopped production to talk with chat about certain decisions and changed focus or method to better ideas that the viewers have. It's been suuuuper cool, and I'd be overjoyed to have more people tuning in to help. A lot of people work on their own projects while watching the stream, so come check it out! There's a link to it in my signature at the bottom of my posts.
Off stream, and part of the reason I haven't posted much lately, I've been really grinding in on the terrace structure itself which has involved a lot of hand-placing grass cards (as shown above), trim decals on the floor, and bending railings around the perimeter. I've also worked a bunch on the main section of the terrace platform over by the statue which is our control point area. I split that area into smaller sections, and did a texturing/decal/detail pass to the metal plate floors in the corners. Also added a light strip trim and lighting to the space:
As a side note, if you're wondering what went into the overall layout of the Terrace, I initially blocked in just a big giant open area, but I quickly realized how awful that was going to be. While brainstorming things that seemed inspirational of Overwatch, I remembered the World's Fairs that we used to do back mid-last-century. Their fairgrounds were my point of inspiration for how to layout my exterior playspace:
And my result as it stands today:
Obviously a completely different scale, but I tried to use similar pathing philosophies to break up the space, sightlines, and drive a player around the space. (By the way, that OW logo is legit just slapped on there and is temporary).
Speaking of scale, that's the final thing I've been really focused on addressing. I got some very, very awesome feedback from a very kind and trusted source, and it cut straight to an issue that's been creeping in to the project more and more: scale. One of the biggest culprits was the West Hallway. Here's an older screenshot:
It will become obvious to probably most anyone when you look from the doorway to the blue rubber anti-fatigue mat, to the 'cables' that the scale here is WAY out of whack. Like, hilariously so. One of the biggest challenges I've had in this project from the beginning has been the question "How chunky?" How thick and exaggerated to I go? Lots of things in OW feel very voluminous and solid with exaggerated bevels and curves, but a lot of other things can get really intricate as well. This hallway is a great example of me trying and failing to serve both of those masters at once. It's fine to make things chunky and solid, and it's fine to go intricate, but everything that shares the same physical space needs to play well relative to each other or it all falls apart.
Another exacerbating factor here is that I've been putting my prop list off until after I finish the architecture. Looking at the above screenshot, there's not a whole lot of things in that you or I could have encountered in our actual irl lives, right? I know doorways, but I have no idea how big sci-fi doorways should be. I have no idea how big those lights are or what size those pipes should be. There's no anchoring, instantly recognizable objects in the scene at all so even if the scale was accurate, the viewer's brain would still be at a loss. The closest thing I have in this shot is that anti-fatigue rubber mat, and it happens to be the worst scale offender in the shot AND it's in the extreme foreground, emphasizing the effect. Yikes.
SO I've been working on a scale pass. You may have noticed a Commader Morrison model in an earlier shot. I was able to find a copy of him online and I brought him into the space to help reset my scale. I've resized various elements throughout the level and I've begun adding some initial props; very obvious, recognizable props, to the space. Here's the after:
I feel it's moving in the right direction now. I may do some other things, for example the white PVC pipes are gigantic. I may shrink them in size, but add more of them, for example. Please do feel free to nitpick me on this. I can't guarantee I'll be able to resize everything but I've also been looking at these spaces for so long that my eyes gloss over a lot these days. Feedback and critique is invaluable!
Here's another area that benefits, even just using the same props again:
Finally, another distraction that's kept me away from the thread is GDC! I'll be attending GDC for the first time this year. I had these snazzy business cards printed up with some familiar artwork:
If you're going to GDC and you'd like to get in touch with me, feel free to message me here! I'm looking to meet new people and start to get a feel for my next career moves, so let's talk!
In the meantime, I'm getting back into a groove on the project again after some distractions and some more tedious tasks, so things are moving forward. Hit me up and check out the streams if you're interested!
Other than that, the only thing that still stands out to me scale wise in the second shot is the red handle near the 'No Access' blue door. It seems too big (granted some of the Overwatch characters have pretty meaty mits so maybe it works in context of the Overwatch universe).
Cool to see the update and have fun at GDC!
In Overwatch textures look very clean - there is not much (if any at all) of small sharp noise from dirt and grime in there. If we see some destruction, sctraches etc its midsized or pretty big in size and if there is any dirt it looks very smooth and gentle. Also they like to use bright and clean colors everywhere and not too many of different "base" colors at the same time. We see balanced and considered color schemes - blue and yellow, yellow and orange, pink and blue, brown and green and so on. And colored lights often help to achive it. I included some screenshots to describe better what i'm trying to say.
Once again - great thread, thank you for sharing and sorry if i'm saying something too obvious!
@Elaisu Thank you so much! You're right. Overwatch textures definitely have a lot of hand-painted information in them, but the primary read is always color first, so much of that detail is allowed to disappear in favor of the overall scene composition, especially from a mid- to long-distance. Thanks for the feedback!
@sybrix Yeah, there's just something about the Maya viewports that short circuit my brain when it comes to scale. Getting the references in there has helped a ton.
As for the stream, if anyone hasn't seen the stream and has any questions or feedback for the environment, come join!
My biggest issue with the shots your posting right now is the overall readability and noise of them. At first, I wanted to place the blame for this on your screenspace AO, but that's not quite it. So I took some of your shots into photoshop and looked at them without color and alot of different offenders raised their heads.
You have your screen-space AO, you have bloom, you have deep shadows and you have metallic materials all contributing to blowing out your blacks and whites. but I don't want to talk about any of that.
You made me really want to just get into Value Ranges. So, with apologies:
Here's the image I started with:
So when we look at just the value of the shot, it makes it easier to call out problematic areas. For me, the big issue is the amount of very dark grey bumping up against middling values and chopping up the picture in a way that doesn't help define the volume of the space. Instead it helps define the assets it surrounds. It's a classic environment art mistake where you want to call out the solid assets you made, but the scene suffers as a result.
I feel like when you re-lit some of these scenes in the past, you where probably aware of this advice and you didn't shy away from making parts of the scene dark, in order to give the overall scene more character, and call out lighter areas. This is good iteration - but it's not there yet.
What we are left with is dark areas of the image that still feel cluttered; the flat surfaces of meshes are in shadow, and the contact shadows between props is pushed darker still to almost pure black. Meanwhile lighter areas have a tendency to blow out, and when these blow outs happen next to dark areas, like in the background of the image above, we get the full value range across a small visual area. All of these contrast points compete for our attention. Finally, the super-dark contact shadows break up the scene in a way that isn't relevant to the space, but to the props they are contacting with.
One of the worst contributors to this is the shiny metal surfaces: black and white values combined. When I looked into how blizzard managed their values, I think they might have had the same issue with metal surfaces, but have cleverly side stepped it. I'll come back to that.
Uh, part 2 I guess:
Lets Look at some screenshots from Overwatch maps that seem closest to your style goal:
So right off the bat there's a clear difference, but just seeing the goal is not particularly useful or we'd all be perfectly nailing our art ever time we opened artstation for ideas.
My first thought is, while the values here might look super clean and crunched, there's probably some post processing happening in the game like UE4's tonemapper that crushes the blacks and softens the highlights. So lets accept that as a final tweak you can apply and move on to the more subtle solutions.
*break to get a beer*
So we know value is important in art, but I feel like the application of value in environment art is often mishandled. Back in the day, you didn't have normal maps and you painted light directly into your model. This led to problems when your model got lit, not just because the lighting might not line up with the shadows you painted, but because if you painted a white highlight there was no way the light could make it brighter. The solution is to paint your values around a middle range and give your lightest values somewhere to go when they get hit directly by a GoldSrc Engine spotlight.
But the year is now and we do have normal maps. And you are controlling the lighting in this environment! You can plan what you want your values to be, and where they go when the light hits them. You can still leverage the idea of working within a value range to create great composition.
What I mean by this is... think about how you can break apart the space of the level into separate value groups that identify the key large shapes in your design. (Your plaza concepts, by the way, are killer). While this can be a design detail (like the floor logo), it's more important that you are describing the volumes of the environment this way. Lean into the level design aspect of environment art because lighting is maybe closer to level design than env art in this regard.
Below I've slapped a surface blur on Nepal and your shot from before:
what I see from the Nepal shot, is clear zones. The ground and the steps. demarking walkable space is key in environments. Obviously players can figure out where they should and should not be able to walk anyway, but you want to offload as much of that thinking from them as possible, and make it intuitive. Most of design advice is like this. It's not about can they figure it out, it's more about making it so obvious that there is no figuring to be done.
Wait where was I, uh, the floor! and then the Buildings, with a subtle but readable breaking up between the building's shape and the decorations that give it a unique identity. Let's dive into that - back to photoshop!
So I've marked the values from my blurred image on the gradient to the right. And then slapped those values next to each other so you can see the contrast they have with each other. The values have enough contrast with each other to be able to describe the identity of a building, but they only occupy 10% of the value range! I know this blurred ass image barely represents the original screenshot but I hope it makes the point clear.
We don't need to see a full value range from black to white in the same way you don't need to use every color in the palette to paint a landscape. You add detail within contrast to provide interest, but that detail can happen at a level of contrast subtle enough that it doesn't demand your attention. When that detail is demanding your attention at the same volume as the very structure of the space it sits in, things get noisy.
I'm really running the risk of muddling though what other people have described much better. To be honest I have typed this out in a stream of consciousness and I'm not sure if this will be useful to you specifically.
So let me wrap this up with some potential solutions of the top of my head:
Finally - while I was looking into this, every single screenshot from an artist who worked on the project has a super high Field of View. Overwatch is a FPS and part of being an FPS is experiencing the game with an 80+ FOV. I'm certain the maps where designed with this in mind.
sorry for rambling. Hope this helps someone...
I think you really nailed some things that have been bugging me visually, especially about the interior, but I'm solidly in that space where it's getting hard for me to see the forest for the trees.
As for the FOV, people who watch the streams can attest, whenever I test the level I always set the FOV to 100 before I run around. That doesn't come through in the regular camera view that I use to take screenshots, though. I should look into how to set that default FOV.
We're gonna talk about environment art in general and stuff I've been thinking about and learning since I last streamed last year. I'll also do some portfolio/project/whatever critiques/feedback towards the end if people come and want!
Anyway, back to it!