Article best viewed in Firefox or Opera internet browsers.
Below is an interview with Ivan Simoncini, who is a level designer and artist at Valve. This first section will contain the interview and the second section will go into more production and design detail.
Kanga: Hi Ivan and welcome to Game Artist. Thanks for taking time out to talk to us. Can you tell us a bit about yourself?
I'm Ivan Simoncini. I've been working as a level designer and environment artist for... gee... almost 14 years. I started out making the usual deathmatch maps, mainly using the Unreal engine. A couple friends and I founded Black Cat Games in 2005 and developed a game called Alien Swarm. It was a very fun experience since we were such a small team. Every decision each one of us made had a huge impact on the product. Valve played the game and liked it. They asked us if we could imagine working at Valve - an offer we couldn't refuse.
Kanga: First off, compliments on the promotion videos for Portal 2. The humour used is dark and very appealing. The overall production looks fresh and vibrant. Were there any original team members from the initial *Narbacular Drop student project working on Portal 2? If there were members of the original crew that created Portal what were their roles and how did they influence the production of the sequel?
Quite a few of the original team were working on Portal 2 - coders, artists, level designers. The Portal 2 team was significantly bigger than the one who developed the first Portal game. When we go to 'shipping mode' almost everyone at Valve is working on the game to some degree. People are usually excited if new members with fresh (and usually crazy) ideas join a team. The luxury of having Portal veterans and noobies on the team was, of course, fantastic.
Kanga: Ok, the levels in Portal 2 are king, being the main mechanism for the game. Were there other major influences that played on artists aside from the previous version? What did the environment team use to draw inspiration from?
We ploughed our way through vast quantities of reference material. Researching is a big and fun part whenever we tackle new environments. We looked at photos from NASA (Apollo area), submarine bases, Chinese apartment blocks under construction, overgrown and destroyed temples, abandoned dry docks, seedy American motel rooms, metro stations, salt mines etc.
Doing color and material studies for the old part of Aperture Science was a lot of fun. We looked at a lot of ads from the 40s, 70s and 80s. It's amazing to see what color combinations they used a few decades ago and how that reflected a certain way of life. It's also stunning how 'political correctness' changed in a rather short time.
You'll notice a certain decrease of colors when you work your way up again to the modern parts of Aperture Science. And the world becomes less extroverted - a reflection of Aperture's financial downfall and Cave Johnson's state of health.
Kanga: A video presentation IGS 2007: Kim Swift - 'Our Journey From Narbacular Drop To Portal' by Kim Swift laid down very solid design principles used on the first Portal. Were there any new approaches you discovered while making Portal 2 that aided you in producing a better game?
I can't state enough how important playtesting is here at Valve. Some of the best ideas (on paper) simply might not work out if you implement them. Our workflow usually is:
1) Make a quick prototype of your idea.
2) Have the guy sitting next to you play it. You'll usually get a lot of constructive feedback.
3) Make changes accordingly to feedback.
4) Have other developers play your levels.
5) Iterate some more.
6) Get external playtesters and watch them play your map.
7) Gather more feedback
Some of your work might never make it past 2).
This might sound gruesome but it's actually a very constructive and satisfying way to work. There's always somebody who can make your work twice as good.
Kanga: What level design techniques were used to reinforce the connection between the player and the game?
We tried to create a very intimate and personal relationship between the three main characters - Chell (the player), Glados and Wheatley. It's mainly achieved through narration, scripting and, of course, timing.
Kanga: There are very strong elements present in the levels, what tricks did you use to stop the gameplay from becoming repetitive?
We're always trying to surprise the player by introducing at least one new interesting element in every map - a new gameplay element, story twist, animation, vista... Very often it's a combination of all the things mentioned above.
It's a lot of fun designing puzzles if you take two elements and mix them together. For instance: So what happens if I take turrets and orange paint?
Kanga: I assume the Valve engine has changed since the Halflife 2 days. How did changes to the engine help your process?
The source engine is very flexible. It's got one of the most powerful scripting systems I've ever worked with.
The introduction of dynamic lights was the best new technical feature from an artistic standpoint. Everything looks almost twice as good if you throw in a dynamic light source. Moving shadows are also very helpful to direct the player's attention to an important moment or spot in the level.
Kanga: Thanks very much for your time Ivan, are there any parting words you have for eager level artists and designers out there.
Get your work out there and be open to feedback. Iterate and finish your level, model, concept art etc. This gives the reviewer a chance to look at your work and the progress you made.
I would highly recommend joining a mod team and ship a game before you apply for a job. Having a finished game in your portfolio means you know what it's like to work under pressure and being exposed to criticism.
*Ed footnote. Narbacular Drop was a nonsensical word that meant nothing. If googled at that time only the game would show up as a result.
Welcome to Portal 2 :More detail from Ivan Simoncini
A good computer game is really hard to make. A lot of people from a variety of disciplines have to come together around a shared vision then execute and - with input from testers - refine that vision until they get it just right. At least, that’s how it happens at Valve; we brainstorm, experiment, then iterate to death, test driving different approaches until the wheels fall off – or not.
The container ride at the very beginning of Portal 2 is a case in point.
Portal 2 begins with Chell, the player’s avatar, waking up in a small, seedy motel room.
The initial plan was to have the room reconfigure itself into a testing chamber, letting players know they were still trapped in the Aperture Science labs. It seemed like an original approach. But in testing, a rundown folding motel room just wasn’t visually exciting enough. More importantly, it didn’t deliver a key message: the opening vista had to convey the fact that the player, along with thousands of other test subjects, had been hibernating in an enormous facility for a very long time.
We experimented with the more traditional sci-fi idea of sterile pods of test subjects awaiting reanimation. The upside was that players would immediately grasp what was going on and we wouldn't have to waste a lot of game time on exposition.
But eventually we agreed that this was too clichéd for the quirky world of Aperture Science. The team also wanted an exciting vista for the opening of Portal 2. The artists delivered a dark, post-industrial landscape where a vast ruined canyon snakes through jumbled stacks of crushed, rusting shipping containers. Sitting inside the containers are the “motel rooms” that house Chell and the other the test subjects.
Early prototype of the stasis chamber
Concept art of the intro space
As an intro to Portal 2, this container canyon solved two problems: it was visually arresting and it also hinted at just how much time had passed since the first Portal game. The container ride would take the player through this physical space which would rapidly crumble away to reveal the real world of Aperture. (It was also a neat homage to the well-known intro train ride from Half-Life 1).
But the ride itself, which moved the player’s container from its original stored location through the canyon and back to the active testing facility, presented a huge challenge. In fact, the idea had been abandoned earlier in the game development process as too complicated and risky. This concept art changed our minds. But we still had to figure out how to do pull off the container ride itself.
Motel room color study
From concept to game
The animators began by defining a path that described the container ride’s trajectory through the space. They gave the path coordinates to the artists who were creating props, materials and textures for the environment. The artists were allowed to enhance everything but this corridor as they began creating the first vista players see as the container walls start breaking away. Artists were careful to avoid placing any decoration in the container’s pathway. Doing so would block the container (and the player inside it).
Prototype of the container ride.
While programmers and animators got busy scripting the container destruction ride the artists started building art assets. We wanted the space to resemble a huge canyon that fades into nothingness and hints at the still larger structures and areas of Aperture that are poking through huge cracks in the space’s walls and ceiling.
We needed a few tricks to realize this vision and stay within memory (budget). For example, we used set pieces that looked detailed but actually consisted of a very small number of polygons. The simpler the shapes, the less rendering time. Some of the containers in the far distance were replaced with basic cubes. Tilting, twisting or otherwise rearranging them lent visual variety. In this way, we were able to achieve a high density of detail with a limited number of props.
First art-pass: structuring the space by using big details, then going smaller and smaller, picking a few spots and make them look really detailed.
Same shot in wireframe view, basically showing that we're rendering LOTS AND LOTS of triangles.
Second art-pass: Adding materials, lighting, and post processing adjustment of color saturation, contrast, etc.
The Art of Chaos
By far the biggest challenge we had to overcome in order to make the container ride work was the player. In this animated sequence, the player is inside the simulation instead of watching it from the outside - something we never did before. The player’s container is crashing into canyon walls, furniture is flying around the motel room, and steel beams and other debris are piercing the walls. The player could get hit, crushed or thrown out of the container altogether. It’s dangerous in there! That’s not good. Here’s why.
A lot of complex math goes into defining any interaction between the player and a physical object. If those algorithms misfire bad things happen: the container ride could just stop mid-air, the game could shut down or the player could die.
To protect the player, we built two containers. One model is the super detailed container you actually see in the game. Somewhere else in the level is a kind of shadow container, a very rough, hollow cube. The actual player is inside the simplified container. The player camera, or the player's viewpoint, is inside the real container. The two-container strategy gave us better control over the simulation, which kept the player, the ride and the game safe.
We created the container’s destruction pass in Maya. We used a complicated mix of real-time physics simulations and handmade key-frame animation so we’d have enough control over all the elements once the mayhem starts. The total compute time for the coarse and fine simulations was more than 15 hours. We went through 96 iterations before we were happy with the result, and some sleepless nights trying to get the simulation to run on PC, MAC, Xbox and Play Station 3 platforms).
The coarse simulation was designed to compute the container’s gross motion. The fine simulation rig reads the coarse cache data and then calculates the destruction of all the interior debris and props.
Pieces of a Puzzle
We were feeling pretty good about the end result, but when we started play testing it we realized that players were overwhelmed by all the things we were throwing at them. They were trapped in a confined space with the world literally crumbling away around them. On top of that we gave them Wheatley chattering and a very intense soundscape. We decided to dial things back a bit until we found a better balance. Basically, we streamlined the amount of information to arrive at an experience that the player could not only endure but enjoy – and understand. We ended up reducing the amount of visual input.
Simplifying the intro sequence gave it more punch, and improved performance on all platforms.
It's safe to say that the intro of the game was one of the biggest hurdles we had to overcome when we built Portal 2. We seriously hope you'll enjoy the ride.