Building the levels: An exclusive interview with Valve level designer Phil Co
Phil Co is a Level designer at Valve, he has helped create some of the most memorable and iconic levels from their award winning series. He has worked on Half-life 2 episode 2, both Left 4 Dead 1 & 2, and more recently the critically acclaimed Portal 2. He is also the author of the essential guide for aspiring level designers “Level Design or Games: Creating Compelling Game Experiences.”
What is it like working at Valve?
Working at Valve has been absolutely amazing. I’ve been working here for 8 years, and the company never ceases to impress me. It is filled with so much talent. I get to work with people who have created my favourite games and even people who worked on some of my favourite films. We have a flat organization where everyone is treated equally, and we are all co-located so everyone has equal opportunities to contribute.
Could you tell us what kind of PC setup you use at Valve? Do you use anything special to optimise workflow, such as a keyboard or mice with macros?
My work station is a PC (i7, Nvidia GTX 680, 32 gigs of RAM) with Windows 7. I have a dual monitor setup where the main monitor has my editing programs and my side monitor has extras such as mail, web browsing, or streaming competitive Dota 2 matches! I just use a mouse and keyboard. I have a Wacom tablet for when I use Maya, but that's only occasionally.
The entire station is on an adjustable height desk with wheels. The only things you need to plug in are the power strip and the network cable. At Valve, everyone uses the same kind of desk with wheels so that we can move freely from project to project or to a different cabal room if we expand the team. I have probably moved at least one time every year since I joined Valve.
What is the usual workflow you follow as a level designer at Valve?
On a typical project at Valve, the cabal (smaller subset of the team which consists of 6-8 people) works on a weekly cycle. Every Friday, we would playtest the section of the game that we're responsible for with someone from outside the team (even outside of the company depending on how far along we are). Everything we do during the week is focused on that Friday playtest.
On Mondays, we have meetings where we decide what the goals are for the playtest and who is responsible for each task. As a level designer, you're the owner of the level you're working on. If you're coming up with a brand new level, then you might call a meeting to design it. If you're iterating on an existing level, then you would work closely with the other members of your cabal to get everything you need for that level. You might have to prototype a new mechanic which would involve a programmer, or you might have to perform an art pass of your level which would involve an artist.
Once production on a game has commenced, how many levels do you usually work on simultaneously?
The amount of levels you would work on simultaneously really depends on the project. On Portal 2, I worked on several test chambers at the same time (some levels were passed on to me because they had to fit into a certain section of the game).
The test chambers in Portal 2 needed to flow together because of the way the puzzles used mechanics from previous chambers and would culminate in more complex puzzles.
I worked on a few of the Excursion Funnel levels while also making Wheatley’s failed test chambers.
Whereas, on Left 4 Dead 2, I would work on a single level on each campaign. On the Left 4 Dead series, we liked to work on campaigns as a whole until they started to feel shippable and then we moved to the next campaign.
How do you decide which levels to build first, do you start chronologically or do you focus on building levels that contain key story elements?
The order of levels to work on also depends on the project. In Left 4 Dead 2, we actually started with "The Parish" campaign (the last one) and worked backwards from there to the first campaign, "Dead Center." We knew that so many of our customers from Left 4 Dead wanted a campaign with a shopping mall, but we just weren't sure how to fulfill that goal at the beginning of the project, so we decided to save it for last.
On Portal 2, we built test chambers based on mechanics first. We had these mechanics (the light bridge, the tractor beam, the catapult, the paint, the laser) and had all these individual chambers that contained puzzles for them. They were all worked on simultaneously by 3 separate cabals on the team.
We had a simple text file that would put the chambers in whatever order we wanted so that we could test the whole game. We saved the very end of the game until the end of the development cycle. I didn't start working on the final boss level until August, 2010.
What is your initial starting point when creating a level? Does a writer give you a brief description of how they imagine the level? Do you work from an artist’s concept? Or perhaps both?
The initial starting point for a level is always a meeting with everyone in the cabal. Valve has several meeting rooms and they always have a whiteboard that is covered in diagrams, sketches and notes. It's impossible to find a clean whiteboard in the whole office!
The level designer calls a meeting in one of these rooms and starts going through what the goals of the level are. Sometimes the level needs to introduce a new mechanic or a new enemy. Sometimes we need to convey a piece of the story arc. We would list these goals and start diagramming what happens in the level moment to moment. A lot of the details are left up to the level designer, but we try to get the high level design of the level clear so that everyone is on the same page and we can build off of each other's ideas.
Have there been any levels you have built that were cut from their respective games?
I have built a few levels that have been cut from the final product. The most recent one was in Portal 2. At one point in the project, Chell was going to meet another personality core (not Wheatley). I designed a level where this core sabotaged a test chamber so that he could communicate with you privately. He opened up the chamber so that you could talk to him in private.
We wound up keeping the idea (Wheatley prevents a door from opening so he can talk to you through an office window) but we scrapped the test chamber with the puzzle.
When developing gameplay how do you decide if an area should contain a certain gameplay element such as a puzzle or a combat sequence, is this decided before an area is built, or do you sometimes design one for an already built area?
We might have an initial idea that we want a certain level to contain a puzzle or combat. We try to look at what comes before and after and keep in mind what we think the overall pacing should be. We learn everything from playtests, however, so we might have too much of one particular gameplay and the playtester suffers from fatigue. We correct this issue in multiple ways.
For example, in Episode 2, where the player gets the car on the bridge, we decided to add a combat high moment. Most of the level up to getting the car is combat but with just a few enemies at a time. It was made to be sort of a funhouse of zombies. We added the building where you pop the roof panels off so that Alyx can help you with the sniper rifle based on our playtesting feedback.
What would you estimate is the average amount of time it takes you to complete a level from its first iteration to its completed state? Is there a level that you were able to take from an initial idea to completion relatively quickly without many major changes and iterations? Furthermore is there a level that took a particularly long time to develop?
During Left 4 Dead 2, we were getting a little behind schedule and we really wanted to create the entire game in a single year. At the time, we were working on "Dark Carnival." I was working on the Fairgrounds level and my coworker, Dario, was working on the Concert level. We needed to get the Coaster map in so that we could start playtesting and Dario and I split up the map. He built the roller coaster track section and I did the Tunnel of Love section.
We had the whole thing ready to test in just a few days and it was very close to how we wound up shipping it. That was probably the shortest time it took to create a level. The longest by far was the final boss level for Portal 2. That took me 4 months! Final boss levels are super hard to make!
How does the artist’s asset pipeline feed into your level design process? Is a library of modular assets created in advance of the level being built?
We have a pretty efficient way of integrating art within the levels. The level designers create the blocked-in version of the level as quickly and simply as possible (while still keeping in mind the setting). We use simple brushes, existing art, and sometimes even make our own models. At the same time, artists are creating style-guides for sections of the game. The style-guides are made so that you can copy and paste elements into your map. Level designers will modify those elements based on the needs of the level and when we move on to another level, an artist can take over the map and do a final art pass. We also will take screenshots at various points in the process and a concept artist will do a paint-over onto the screenshot to add details, textures, and other features to really make it look beautiful.
The turnaround time for Left 4 Dead 2 was particularly fast, during this time you worked on 4 of the 5 campaigns that shipped with the game, what were the challenges in completing these campaigns within the given timeframe?
On Left 4 Dead 2, we knew the gameplay that we wanted to create, but we were challenged with expanding that gameplay into unique experiences throughout the game in a short amount of time. For example, we thought about all of the fantasies people have in their heads about a zombie apocalypse and one was an amusement park.
We listed all of the elements that make an amusement park iconic and unique and tried to incorporate all of those ideas into the classic Left 4 Dead model. So, playing carnival games, going down the big slide, running on a roller coaster track, and holding out on concert stage were all challenging settings to fit into the schedule, but we knew they would pay off in the end.
Levels for the original Half-life consisted mostly of brush work and few models, today hardware is much more powerful and as a result levels are increasingly becoming more model based. Do you predict an end to brush built levels in future valve games?
I don't think brushes will be obsolete in the near future. They are just the quickest way to build levels and test gameplay. It might be true that more of the final art will be models, but to get us to that stage, we need brushes.
The visual jump between Portal and Portal 2 is huge, was it challenging recreating areas from the previous game in higher detail but still keeping the familiarity?
Portal 2 really benefited from concept art and style-guides. All of the test chambers we designed actually looked the same (very similar to Portal 1) at some point in the development cycle. After the puzzles were designed, we grouped the levels into sections based on mechanics. Once that was feeling good in testing, we started to incorporate the visual differences that would establish the art style of Portal 2 and distinguish the test chambers from those in Portal 1.
Level designers could even help out other cabals by working on the art style on another part of the game than what their cabal was responsible for. So, even though my cabal was responsible for the last section of the game, we would take test chambers from the first section and dress them up.
Which is your favourite Valve universe to design and create levels for and why?
Portal 2 is probably my favorite Valve universe to design levels for. The main reason is that anything goes. You can base your design on modern reality or go in a completely different direction where everything is pretty crazy. Test chambers could be whatever you wanted them to be, but when you went "behind the scenes," there was this realistic structure and method to the madness.
Out of all the Source Engine levels you have built, which are you the most proud of and why?
The one level I'm most proud of is the final boss level in Portal 2. Among the level designers at Valve, we have this running joke that each one of us has a duty to work on a final boss level in a game. After that, you have fulfilled your duty. It was incredibly challenging for me to try to match the ending of Portal 1.
At the time, Portal 1's ending was listed as among the best in any video game. Starting with such high expectations didn't make things easier for me. It did take me a very long time (the longest I've spent on any level), but I think it turned out pretty well and it wound up being a very rewarding experience. So many members of the team contributed to the end of Portal 2, and I got to work with all of them. I learned a lot.
In the making of Portal 2 it is revealed that early in its development the game was taking a different direction, did you work on this version and if so is there anything you can reveal about what the levels were like?
Portal 2 went through many versions (including a time when there wasn't a Portal gun), so I'm not sure which version you are asking about. I can say that we cut a lot of the ideas we had over the course of the project and wound up putting many of those ideas back into the game in some form. At one point, there were other personality cores other than Wheatley who would interact with you. We cut them from the game and wound up bringing some back for the final boss fight.
There was a time when the beginning of the game was walking through a diorama of scenes where you met Cave Johnson's voice. The art style of these dioramas and Cave Johnson got cut and then put back in for the underground section of the game. We always felt that if we missed something enough, it would make its way back into the game.
Portal 2 had a simultaneous release on PC, Xbox 360 and Ps3 Is this something that provided any challenges whilst building and designing the levels?
On Portal 2, we designed everything with the gameplay in mind first. There were parts we had to optimize so that it would run on all platforms, of course. The most notable sections were in the underground section with the huge vistas and the existence of paint. We would sometimes replace water with bottomless pits.
At the end of the game, when you are traveling between test chambers by tractor beams instead of elevators, we also had to make some performance compromises, but the overall experience stayed the same.
You designed the epic boss fight that takes place at the end of Portal 2. Could you talk us through the development of this level?
For the final boss level in Portal 2, we started out with just a few goals:
- Simple enough that players would not become stuck
- Incorporate multiple kinds of mechanics to give the player a sense of complexity
- Have stages of progression that could change the environment
- Remind players of the Portal 1 ending
For the actual battle, a concept artist named Jeremy Bennett created some images of Wheatley with iconic pieces of Aperture Science attached to him. The idea was that he knew you were coming to defeat him and he wanted to protect himself in this very haphazard way. The first concept we pursued from these images was Wheatley with a bunch of turrets attached onto him.
The first prototype I created for the level was Wheatley with these turrets attached and you had to launch a cube at him using catapults aimed at him to knock the turrets off one by one. When you knocked the last one off, you would hit him with the cube again and defeat him. We didn't know what the cube was going to be and we didn't know if you would use catapults to launch them. We just wanted to test if the mechanic was fun to hit him with things. This also would remind players of the Portal 1 ending where you redirected rockets to hit Glados. However, we discovered that the turrets on Wheatley were too punishing.
We tried having shield panels coming up from the floor and out from the walls to hide and time the launches, but players became frustrated if they missed because they would take so much damage from the turrets shooting them.
We decided to replace the turrets with the shield panels - as if Wheatley specifically guarded himself from projectiles. An artist named Alireza Razmpoosh created some tests with the shields animating and protecting Wheatley.
We also decided to incorporate the most impressive mechanic in the game, White Paint. Because Wheatley would do everything to stop you from defeating him, he would remove portal surfaces from his chamber but he would forget about the white paint. The first part of the level would be to get Wheatley to break a pipe to expose the chamber to White Paint. We tried a few different ways he would break the pipe, but decided he should break it with the same weapons he tried to use against you - bombs. Redirecting bombs that Wheatley lobs at you to hit him with portal placement proved to be extremely fun. It reminded players of Portal 1's rockets but was different enough to be new. Combining this with the White Paint fulfilled our goals of seeming complex. Garret Rickey made me a bomb shooter and Richard Lord added the cannon to Wheatley’s model.
When you hit Wheatley with a bomb, we wanted to try a reversal of Portal 1's boss fight. Instead of removing cores from Glados, you would be attaching defective cores onto Wheatley. We even had personality cores that got cut from the game we could use. This was the next prototype I built was a simple level where cores would get dropped into the level and you would use different mechanics to plug them onto Wheatley.
- An early prototype of the chamber where you would need to use multiple mechanics to attach cores to Wheatley.
Every time you placed a core onto him, he would change his position and you had to figure out how to get the next one on him. The first was easy. You just placed a portal on a wall next to him and you could place the core onto him through the portal. Then you had to place a portal high on a wall and come out of it to land on the bounce paint to place the second core. There was a speed run into a portal to come out of the floor and pop it onto him as the last core. All of these seemed pretty fun to do, so I integrated them into the level where you hit him with bombs. It turned out to be too challenging. So, I tried to reverse the situation a bit again, and this time the player would do these maneuvers to get to the cores. Placing them onto Wheatley would be trivial. This proved to be much simpler and Wheatley could remain in the same position in the chamber. Now, we could start shaping the chamber.
Since we wanted the fight to be in stages with the environment changing at each interval, we had the chamber slowly be destroyed revealing different paint. The paint was necessary to get to the next core. At first, we tried getting the player to destroy parts of the chamber. You used to have to portal up to a catwalk and get Wheatley to try to hit you with a bomb there so that he would destroy the catwalk and a pipe below it exposing bounce paint. There was another pipe that you would have to redirect a bomb to exposing speed paint when it broke. Once again, this proved to be too convoluted and the simplest fix was that you would destroy the bounce paint pipe when you walked on it and Glados would destroy the speed paint pipe when she delivered the final core to you.
There was a lot of iteration to get the player not just to leave portals in the same positions for the entire fight. This required the bomb shield configuration to be different for each section and for you to have to place your portals to reach the cores. Alireza changed the animations of the shields many times for me while adding some great touches like Wheatley banging the ground underneath him if the player got caught there. Mike Belzer also animated Wheatley for his position to shoot the bombs as well as his talking scenes.
Erik Wolpaw and Jay Pinkerton were the writers on Portal 2, and they really wanted this element of comedy towards the end with a stalemate button. We wanted to call back to the button you used when you replaced Glados with Wheatley in the first half of the game, but this time, Wheatley thinks he's outsmarted you again.
The writers wanted to have a button surrounded by bombs and even a sign that said "Press this." We eventually decided that you want players to try to press the button without fearing death, so the bombs came out as a surprise (with Wheatley using his evil laugh again). Originally, the comedy continued so that after the button is destroyed, the computer said, “Stalemate button has been destroyed. To initiate core transfer, simply say ‘Yes.’” This was the final thing the player would do and it would stump people at first. They would think “Well, now I’m screwed.” It was funny for us on the team, but didn’t seem like a viable solution. Besides, we never actually got a voice actor for Chell.
In Portal 2’s developer commentary you mention “Playtesters would shoot the moon and instantly turn away, thinking nothing had happened. They didn't realize they actually had shot the moon. We tried and rejected a few different approaches to communicate the effect.” Can you elaborate on what the different approaches were that you tried?
For a long time on the project, there was joke about shooting a portal on the moon. At one point, there was an idea where you would see the moon and if you shot a portal on it, it would suck you out and kill you. We all thought this was funny, but it would teach the wrong lesson. We decided to resurrect this idea for the ending of the game. Since we knew the ending of Portal 2 had to be epic, we thought shooting a portal on the moon was just epic enough.
The key thing to achieve was to have the portal shot at the moon be the last one that you place. This meant that there already had to have been a portal placed underneath Wheatley. The stalemate button room provided the perfect vehicle for this experience as the player would have to place a portal under Wheatley to get into the room. A fire in the chamber caused a sprinkler system to clean up all the other other portal surfaces except the area under Wheatley (which was blocked by him) and the player had to use that spot. After the stalemate button is destroyed, the player would be knocked back and look up from the ground where part of the ceiling would fall away revealing the moon.
We started testing the ending with the moon and it was an immediate success. The problem we had, however, was what happened after Wheatley gets sucked through the portal. Andrew Burke and Matthew Russell were the animators tasked with getting Chell and Wheatley sucked through and showing how Chell gets back into Aperture Science.
Originally, Chell was supposed to have her leg tangled in cables that dangled from Glados' body as she held onto Wheatley upside down. She let him go because the force of the suction was so powerful and the cables automatically reeled her back in. When we asked playtesters what happened, they told us that Glados had saved them. We wanted to make this even more obvious, so we added the robot arm and Glados' voice.
The ending scene with the grass fields is rendered in Source Filmmaker, what led to this being pre-rendered and not a scripted in-game event?
It was also a decision early on that Chell would be freed because Glados didn't want her any more. The writers wanted her to feel like she got kicked out. This is why the door slams shut behind her and the companion cube gets shoved out after. To get to the surface, there had to be a long elevator ride to last the length of a new Jonathan Coulton song, so I prototyped the speech that Glados gives you where she replaces you with the co-op bots and sends you to the surface in the elevator. She originally said something like "Look, I've found test subjects that are much more reliable than you." We later decided to keep the focus on you and just show the co-op bots when you woke up. As for the elevator ride, we started playing with lining the ride with turrets so that you weren't sure that Glados really wanted you to live or not.
This led to the turrets singing to you as you left which also changed the placement of the Jonathan Coulton song. After the elevator ride was mocked up in Hammer with the proper length and the new turret song, Keith Lango made the elevator ride into a full animation with all the turrets singing. Jason Brashill created the environments including the outdoor surface with the wheat field and the shed. Since we still wanted the player’s last action to be shooting the moon, we decided to just make the elevator ride and the surface an animation. It would also keep the timing perfect between the song and the visuals.
Can you tell us anything about what you have been working on since Portal 2?
After Portal 2 shipped I have been taking a break from level design. Valve is a truly unique and amazing company to work at because you can decide what you want to work on. Of course, you should be making decisions that you think will benefit the company, but a very interesting opportunity came in the summer of 2011, when I was asked to film our Dota 2 tournament in Germany called The International. Together with an artist, Jeff Unay, we started filming what would eventually become a feature-length documentary film “Free to Play” which will be coming out soon.
- The movie poster for Free to Play.
It’s been an incredible experience – filming all over the world (China, Singapore, Ukraine, Denmark and the US), meeting professional players and their families, editing chunks of the movie, and learning all the aspects of filmmaking.
I have also filmed the other International tournaments here in Seattle and have created several videos including the 2012 recap video and the team stories for all the teams that have attended.
We are almost done with the film though, so it will soon be back to level design for me!
Interview by [Steve]