Portal developer commentary

From the Portal Wiki
Jump to: navigation, search

Portal developer commentary.

Developer Commentaries are recorded audio commentaries about the creation of Portal, that can be found once the test subject has completed all the tests and finished the game.

Developer commentary mode

The developer commentary mode is a game mode where the commentaries can be found as floating speech bubbles in several rooms of the Aperture Science Enrichment Center. After finishing the game, the developer commentary mode gets unlocked in the main menu. The commentaries are activated by pressing the Use Item action key on them, and can be stopped in the same way (note that some commentaries can alter the visuals or take temporary control of the test subject while they are playing). Take notice that the test subject can't get any achievement while testing in developer commentary mode.

Commentary text

#000

Creating the AI voice was a multi-step process. First, we ran every line of dialog through an automatic text-to-speech program. In the studio, we cued the actress who plays GLaDOS, Ellen McClane, with the computer-generated sound file. She'd mimic it, and then, over the course of several takes, adjust her performance to clean up any words that were unintelligible in the computer version. For instance, here's a line as Ellen delivered it: <INSERT RAW ELLEN LINE>. Once the recording was done, we processed all of the dialog to give it an extra computery edge. Here's that same line as it appears in the game, with the pitch constrained, pitch modulation suppressed, and the formant moved up: <INSERT LINE HERE>.
— Erik Wolpaw

#001

We designed the post-escape levels to give players brief glimpses of the inner workings of the Enrichment Center. This particular area exposes part of the Storage Cube Distribution System. These little vignettes help make this section of the game visually distinct from the testing chambers while reinforcing the idea that players are now behind-the-scenes.
— Kim Swift

#002

One bizarre fact of game design is that, without some serious prompting, players will rarely look up. In this case, the prompt is a ladder. Most players will investigate where the ladder goes, which is up. The ladder actually falls apart as soon as it's touched, but by then it's served its purpose.
— Jeep Barnett

#004

Through playtesting, we discovered that fatigue set in if we didn't break up the more complicated, deliberately paced puzzles with problems that required the player to perform a much simpler task under time pressure. These pistons proved to be a good foundation for that simpler type of puzzle. Because they feature a moving surface, they also gave us an opportunity to employ some unusual portal transitions, such as ceiling-to-ceiling.
— Garret Rickey

#005

During development, we'd often run across some piece of level design that would break the portal system. For instance, this ceiling-to-ceiling transition was an unexpected edge case that ended up requiring a lot of effort to make work. We tried never to take the easy route of simply changing the level design to work with existing portal technology. We knew that once the game was out in the wild, custom map creators would stress the system in ways we hadn't even considered, so we made it as flexible as possible.
— Paul Graham

#006

An important design goal of these post-escape levels was to give players the sense that they were running wild inside the guts of the facility. With that in mind, we added plenty of nooks and crannies for players to crawl into or place portals through. This ended up being a tricky lighting challenge, as the areas needed to be bright enough to navigate while still being dim enough to feel like obscure, disused sections of the building.
— Greg Coomer

#007

These rocket sentries originally fired lasers. We switched from lasers to rockets after we introduced the glass-breaking mechanic, mainly because glass shattering in a massive rocket explosion turned out to be a lot more satisfying than glass just slowly being melted by a laser. Originally, they also spoke, just like the bullet turrets. Because players often redirect the rockets with their backs to the turret, however, a distinct, uncluttered sound cue was required. The voice simply added too much noise to the rocket redirect mechanic.
— Bill Van Buren

#008

Breaking this tube gives players a chance to test out their newly trained rocket-redirecting and glass-breaking skills in a slightly different context, which helps cement the training.
— Realm Lovejoy

#009

As the end of the game draws closer, we increase the ratio of time-pressured puzzles, like this turret ambush, to give players the sense that they're approaching a climactic encounter.
— Scott Dalton

#010

Originally these catwalks were decorative, but playtesters consistently thought they were significant and often spent a lot of time trying to reach them. We didn't want to stand between people and their desire to walk on the catwalks, so we redesigned the area to make catwalks not only accessible but also necessary to proceed.
— Randy Lundeen

#012

This massive turret ambush was originally a lot more massive, with turrets that dropped from the ceiling and popped out of surprise hatches. In fact, for a while, it was the game's climactic battle. Through playtesting, we learned that this type of pure combat experience, didn't really fit with the preceding few hours of player training. Over several iterations, we toned down the combat and made the room more about using portal momentum to fling yourself great distances, a skill that playtesters really enjoyed using and that's a key component of what eventually became our final battle.
— Jeep Barnett

#013

GLaDOS, the rogue disk operating system that now runs Aperture Science, went through a bunch of design iterations. Earlier versions included a floating brain, a sprawling, spidery mechanism, and an upside-down version of Botticelli's Rise of Venus built out of robot parts and wire. Evenutally, we settled on a huge mechanical device with a delicate robotic figure dangling out of it, which successfully conveys both GLaDOS's raw power and her femininity.
— Jeremy Bennett

#014

The fiction behind this red phone is that, while GLaDOS was being developed, it was somebody's job to sit by it, and, if it ever looked like the AI was becoming sentient and godlike, that person would pick up the phone and call somebody to come help. At the point in time where the actual game takes place, it's become obvious that the Aperture Science Red Phone plan didn't 100% work out.
— Erik Wolpaw

#015

One of the things we learned from Narbacular Drop, our student project that became Portal, was that beginning players often thought portals took them into other spaces or even other dimensions. To help fight that notion, we start players in a visually unique room with memorable objects, so that when they walk through a portal for the first time, they have a clear point of reference, which communicates the idea that they're still in the same basic location. For instance, the radio, which is playing an instrumental version of 'Still Alive,' helps as well by providing some audio continuity
— Kim Swift

#016

Portal is effectively an extended player training exercise. We spend a huge portion of the game introducing a series of gameplay tools, then layering these tools into increasingly difficult puzzles. This layering starts here, where we train the button/box mechanic before introducing the more complicated concept of portals.
— Robin Walker

#017

Welcome to insert game name, here. Hopefully you have already had a chance to enjoy the off-beat perspective and new game mechanic of Portal. To listen to a commentary node, put your crosshair over the floating commentary symbol and press your use key. To stop a commentary node, put your crosshair over the rotating node and press the use key again. Some commentary nodes may take control of the game in order to show something to you. In these cases, simply press your use key again to stop the commentary. Please let me know what you think after you have had a chance to play, as we think we are just at the beginning of taking advantage of this type of gameplay. I can be reached at gaben@valvesoftware.com. Thanks, and have fun!
— Gabe Newell

#018

We put the player character in an orange jumpsuit to reinforce the fact that she's a test subject. Visually, the warmer orange colors help her pop out against the colder tones of the environment. Some playtesters were wondering why she could fall so far without getting hurt the way she would in Half-Life 2. In response, we added mechanized heel springs to her lower legs; afterwards, there was no longer any question about why she could survive such long falls.
— Bay Raitt

#019

These frosted-glass observation rooms make the player feel as if they're being watched at all times, while keeping the identity of these watchers a mystery. The rooms serve a practical purpose as well, since we often use them as convenient and logical light sources for the test chambers.
— Realm Lovejoy

#020

The combination of the prtl-destroying fields (which we call 'fizzlers') and the elevators serve a dual purpose. They provide a clearly identifiable end-point for each test-chamber, while also addressing the more practical problem of how to keep players from portaling across level loads. We eventually integrated the fizzlers into several of our puzzle designs.
— Garret Rickey

#021

It's absolutely critical that players quickly wrap their heads around what a portal is. We noticed early playtesters grasped the concept much more quickly when they caught a glimpse of themselves through a portal. So we deliberately positioned this first portal to ensure that players will invariably see themselves.
— Kim Swift

#022

We wanted players to feel safe while standing in a portal, so we never kill them or destroy objects within a portal that's closing. Instead, we either push or teleport objects out of a portal as it closes.
— Paul Graham

#023

Because we very deliberately introduce and train each gameplay concept, once players reach this spot, we're confident that they know what a portal is and roughly how it works. Early versions of Portal let players stumble through the beginning of the game without always understanding what was going on, which really compromised teaching new concepts. The puzzle you just finished was designed so that stumbling around will virtually always lead to a dead end. Completing the puzzle requires walking through a minimum of five portals in a specific order. This kind of gating, in which a solid understanding of key gameplay concepts is required for success, helped standardize the learning curve of the game tremendously.
— Robin Walker

#024

In early versions of this map, playtesters would charge down the stairs without noticing what was creating the portals. We introduced a mandatory pause in the action - what we call a gate - to help ensure that players stop and notice the portal gun making a blue portal. A particle effect and a loud noise help draw their attention.
— Jason Brashill

#026

This room was designed to make players understand that entrance and exit portals aren't tied to the color of the portal. Playtesters often assumed that orange portals were exit-only, so we created this puzzle to force players to enter an orange portal.
— Kerry Davis

#027

When rendering the player's view through a portal, we must render a separate image using a virtual camera which looks out of the opposite portal. To obtain a correct image and efficient rendering performance, we render only what is visible through the limited field of view of the opposite portal and exclude objects which lie between the virtual camera and the plane of the opposite portal.
— David Kircher

#029

Early versions of Portal featured more detailed, cluttered environments, much like Half Life 2. We quickly realized that unnecessary objects scattered all over the place distracted players to the point where it actually interfered with the portal training process. So we simplified the art style to favor clean, focused test chambers. The modular approach we settled on makes it look plausible that the chambers can reform dynamically on these pistons.
— Nick Maggiore

#032

For training purposes, there's generally just one correct solution to these early puzzles. The original version of this room didn't have the glass barrier. Playtesters would often stand on the button to open the door, and then shoot a blue portal through the opening, bypassing the box entirely. Since this puzzle was meant illustrate the relationship between boxes and buttons, that solution, while clever, was a failure, so we added the glass barrier to prevent it. Later in the game, however, the puzzles become more open-ended.
— Chet Faliszek

#035

Integrating portals with the Source engine's physics system was a complex process that required several iterations to achieve the right balance of performance and correctness. Because portals can be placed virtually anywhere in the game environment, the physics system had to be modified to allow dynamic changes to its representation of colliding geometry, such as the walls and floor around this box, and any objects which may lie on the opposite side of the portal. Initial implementations of this dynamic collision generation system could take up to one half of one second, or 500 milliseconds, to compute the correct collision. This may not sound like a long time in everyday life, but this pause during portal creation was quite noticeable in the context of the game. Ultimately, we designed a system that creates temporary hybrid physics environments in bubbles around the portals using less accurate collision than that produced by Source's standard collision generation, but was accurate enough in practice and reduced the time to create the dynamic collision representation from 500 milliseconds to just 10 milliseconds, which is an imperceptible pause during portal creation.
— David Kircher

#036

For the first few months of development, we rendered the views through portals to two offscreen textures. This approach was easy to implement and was compatible with a wide range of graphics hardware. Unfortunately, this method was incompatible with antialiasing and consumed a large amount of video memory in order to handle recursive views through several portals. Because of these disadvantages, we switched to a system which renders portal views recursively into the frame buffer with the aid of the stencil buffer to isolate pixels corresponding to a given portal. This is a more effective scheme because it is compatible with antialiasing and does not consume any additional video memory for offscreen textures.
— David Kircher

#038

To make puzzles deeper than just teleporting to the exit, we had to include surfaces that won't hold a portal, which are formally introduced here. We experimented with several surface designs before we settled on this one, whose visual noise and reflectivity make it easy to identify at a distance.
— Randy Lundeen

#040

Even though layering player training was a design goal from the start, we still ended up introducing some concepts too quickly. For instance, this used to be the first energy-ball redirection puzzle. Playtesting revealed that this puzzle introduced too many new concepts at once, which ended up frustrating a lot of playtesters. In response, we inserted two test chambers before this one to make the energy-ball redirection training more gradual.
— Chris Chin

#041

Because our test chamber environments were simplified for training purposes, we created visual hot spots within the rooms to guide players' attention. The design is essentially a balance between round objects and sharp objects, the sharp objects representing background elements and the round objects - such as doors and moveable props - comprising our points of visual interest.
— Paul Graham

#043

Originally, these scaffolds ran on electrified tracks. But, crafty playtesters would hop along the rails to the exit, bypassing the puzzle entirely. We tried to solve this by killing players as soon as they touched the rails. That solution ended up being too much of an over correction, as even skilled playtesters were getting frustrated by these one-hit kills, in the more complex puzzles later in the game. Making the scaffolds run along immaterial beams of light solved both problems.
— Jeep Barnett

#044

We previously talked about how we handle static portal collision. But collision with moving objects on the other side of the portal is a completely different and equally hard problem. Walking onto this scaffold was a very iffy proposition for the first few months of development. We solve the problem of colliding with these dynamic objects by cloning the objects from one portal to the other and strictly controlling what objects are allowed to collide with each other and how they're allowed to collide.
— David Kircher

#048

Portal momentum ended up being the hardest concept to convey. For this series of puzzles, which went through more design iterations than virtually any other part of the game, we introduce the idea of redirecting your momentum using portals slowly, step by step. We even have the AI voice pretty explicitly explain the elements of the puzzle, something we avoided throughout most of the rest of the game.
— Garret Rickey

#051

Originally, these exit portal surfaces were static geometry in the final position, but playtesters stubbornly refused to look up to find them. This is another example of the classic game-design problem of coaxing players to look up. By putting the portals on moving pistons, we were able to start them in a position that players were more likely to see.
— Bill Van Buren

#053

Using gravity to fall into one portal so that you come rocketing out the other portal - a skill we call 'flinging' - was another difficult concept to train. We designed a specific visual cue - a pushed out concrete block above a pit with checkerboard concrete tile - to indicate to players that it's time to use the fling maneuver. Repeated several times, this cue helps players associate pushed out concrete slabs with flinging, in much the same way that players learn to associate cubes with buttons.
— Greg Coomer

#054

We found that, rather than looking into a portal to see where it went, playtesters would often leap blindly to their doom. In response to this observation, the moral of this puzzle is 'look before you leap.' The safe orange portal is out of the player's view from this balcony. That forces them to peer through a portal to see it, which trains players in the remote-viewing capabilities of portals.
— Scott Dalton

#056

This room is designed to build anticipation for the big moment when players finally get the fully powered portal gun. The puzzle path brings you in a circle around the device, so that it's virtually always in sight right up until you grab it.
— Lars Jensvold

#057

Player training doesn't always stick, especially after the introduction of a big, new concept. For instance, after they'd acquired the fully powered portal device, playtesters often forgot about the fling maneuver. Since it's such an important skill, this puzzle is designed to reintroduce the idea of flinging.
— Chris Chin

#058

Originally a weighted button was used to open the far door, but playtesters so strongly associated boxes with buttons that they'd get stuck searching hopelessly for a box. We changed the big button to a pedestal mounted push button, thus removing the box association, but playtesters still had trouble realizing they needed to shoot a portal through the door. Adding a ticking timer sound when the door was open cued players that the puzzle expected them to act during that time, which solved the problem.
— Jason Brashill

#059

When we moved from largely placeholder art to our final visual design, this was the first level to get a facelift. We chose this map because it had many of our gameplay elements integrated into a relatively small space. The test chamber art direction was designed to make everything appear purposefully placed. The simple design helps focus players on the puzzles. It also provides a nice contrast to the later, much less sterile behind-the-scenes environments, which contributes to a clear sense of progression.
— Jeremy Bennett

#060

A problem we came across with virtually any puzzle involving boxes and doors was that players could portal the boxes to the other side of the door, thereby trapping themselves in a room with buttons but no boxes. We set up special triggers to detect and handle these cases, and then added the box delivery tubes to ensure players could always be supplied with the required tools.
— Kerry Davis

#061

This is a great spot to appreciate the recursive nature of portals. If you place a portal on each side of this hallway, you'll notice the portals seem to go on forever, similar to the effect you get in a hall of mirrors. In actuality, we support a maximum of nine recursive portal views down any chain of portals. We achieve the impression of infinite recursion by copying part of the previously rendered frame onto the final portal in the recursive chain. It's not absolutely perfect, but it's inexpensive and effective.
— David Kircher

#063

This is the first map in which we experimented with solving the puzzle in as few portal placements as possible. We tried to fit that concept into the story mode, but were never quite able to sell it. Instead of abandoning the idea altogether, we added the concept to a series of post-game 'Challenge' maps.
— Nick Maggiore

#066

A few playtesters put a portal on the floor here and used the rising stair pit to skip the rest of the puzzle. We'll usually rework a level if playtesters discover a way to bypass chunks of the puzzle too easily. But in this and a few other cases where skipping ahead arguably takes more skill than solving the puzzle properly, we let the ninja solution stand.
— Jeep Barnett

#068

We designed this room to draw the player's eyes to the box. The light from the observation room casts horizontal shadows that point at the box, which is directly lit by a warm light from the ceiling. This warm light helps the box stand out against the predominantly cool test chamber lighting. The varying sized squares of the off limits surface also help direct the player's attention upward.
— Lars Jensvold

#069

This particular section went through several iterations because it's the first place where players are required to perform possibly the trickiest portal maneuver, the double fling. Originally, the room also featured an energy ball redirection puzzle. Combined with the new double fling skill, however, it proved to be too much for most playtesters. Overwhelmed players tend not to digest new information, so we simplified the puzzle to require only the double fling. To nudge players in the right direction, we also included visual hints, such as the previously introduced single fling theme - the pushed out concrete slab with a checkerboard landing zone.
— Kim Swift

#070

We noticed through our playtests that many of our players weren't aware of the fact that you could place portals while falling. Since placing portals as you are falling is essential to accomplishing flings and many other portal tricks, we implemented this puzzle, so that the only way to solve it, was to use the mid-air placement tactic. It took numerous iterations and lots of tuning before this consistently played well.
— Garret Rickey

#071

In this puzzle, we wanted players to realize that, using portals, they can travel long distances in a short step. Originally, we had stairs from the ground level to the upper rooms. Many playtesters, however, were able to run from one room to the other quickly enough to consider it a solution without ever thinking to use the even easier portal method. By changing the stairs to very slow moving lifts, we were able to level the travel time for all player experience levels, thus making the intended solution more clear.
— Realm Lovejoy

#072

This puzzle requires players to redirect an energy ball through multiple portal placements. The multiple steps caused many early playtesters to panic and accidentally redirect a ball right into themselves, with deadly results. We redesigned the test chamber so that portals can only be placed in the top half of the room, which solved the problem by generally leaving players out of the ball's reach.
— Paul Graham

#074

Perhaps the most important purpose of this final test chamber is to cement the double fling maneuver training. For previous flings, we employed a visual cue - a pushed out a section of wall. In preparation for the more freeform behind-the-scenes levels, however, we take the training wheels off. To proceed, players must realize that they can double fling almost anywhere they want to, with or without hints.
— Robin Walker

#076

For Portal, we wanted to create a turret that was different from the traditional HL2 turrets. The HL2 turrets are nothing more than mindless mounted guns, in Portal we wanted the turrets to have personality, to be characters. The first step was a redesign of the look. Once we had the look of the new turrets, they seemed less like menacing machine guns and more like cute robots. And robots can never really be cute until they are talking robots. So the next step was creating the vocal characteristics of the turrets. Working with Ellen McClain, we settled on an innocent sounding voice with dialogue to match. This juxtaposition of killing machine and innocent non-aggressive personality ends up making the Portal turret a memorable character.
— Chet Faliszek

#077

Even though Portal tells a simple story, we created a lot of backstory - for Aperture Science, for its employees, for its rivalry with the hated Black Mesa, and for where all of this fits into the cosmology of Half Life. This first Portal game doesn't reveal all of it, but we crammed a lot of little details into the environments. This area, for instance, called the Ratman Den, hints that there may be other people trapped in the facility.
— Erik Wolpaw

#080

Even though we didn't want Portal to be too combat heavy, we wanted to add at least a light combat mechanic. After a few less-than-successful attempts to actually realize this 'light combat' concept, we settled on turrets as a good compromise. The turrets can be defeated with tricky portal maneuvers, but they're also susceptible to a more straightforward berserker approach.
— Realm Lovejoy

#083

The awful end of the companion cube serves a dual purpose: it adds a lot more sinister character to our already pretty sinister AI while simultaneously training players to use the incinerator, a key component of the final level. The training has a nice dramatic payoff, since players later get to avenge the death of their good friend the companion cube by stuffing some of the AI's important parts into exactly the same type of incinerator.
— Jeep Barnett

#084

We designed this hall so that players had to use a box as a shield, but many playtesters read it as an 'avoid-the-energy-ball' timing puzzle, and simply left the box behind. We solved this, in large part, by having the AI talk about the box. A lot. Once the dialog went in, playtesters went from routinely abandoning the box to never wanting it to leave their side. Since we weren't going to make players lug a box around for the rest of the game, we added a scene at the end of the level where players are forced to kill the box.
— Chet Faliszek

#094

Sometimes little details make a huge difference. For instance, we had to add the left side to this wall to give players a visual representation of how thick it is. Without it, playtesters frequently thought the hall had just gotten impossibly narrower, panicked, and stopped making rational decisions.
— Paul Graham

#095

Before the player escapes the fire pit, the AI dialog is all delivered in a computerized monotone. After the escape, however, GLaDOS gets progressively more expressive until, by the end, she's cycling through a whole mess of emotions. The role required an actor capable of mimicking the computer-generated voice while still infusing it with some real character. And since we planned to end the game with a song, she also needed to have a good singing voice. The woman we cast, Ellen McClane, is a great mimic, a terrific actor, and a classically trained operatic soprano. So that worked out pretty good.
— Erik Wolpaw

#096

After players escape, we let them solve an old chamber in a new way. This reintroduction of a familiar space, which can now be solved in an 'incorrect' way, helps convey the sense that players are cheating the system while forging their own path through the facility.
— Realm Lovejoy

#097

We use these moving platforms to add time pressure to puzzles. A sequence of simple decisions feels much more climactic under time pressure, in contrast to most puzzles in Portal which can be solved at the players' leisure. Puzzles that aren't time bound can be more complex, but at the expense of being less dramatic.
— Garret Rickey

#098

In these new areas, we weren't using the surfaces the player is familiar with. The rules for which new surfaces portals can be placed on and which surfaces are off-limits, needed to be taught. So we created this area both to show the player they have escaped into the 'guts' of the facility and to re-train them in which surfaces are valid for portal placement.
— Kim Swift

#101

I know that I was told that, you know, sort of the idea of Valve is that everything... Everything is planned. Everything is worked out in advance. But that's a big, fat lie. Because: The initial idea is started, and they bring an actor into the studio and they say, you know, 'Say these words for us, say this computer generated voice for us.' Okay, I'll do that. Well, then, that gives the creators other ideas. And it's... It's a rolling process; You roll from one idea into the next idea.
— Ellen McLain: Voice of GLaDOS

#102

In the beginning, when I came in, I was just supposed to recreate the sound of a computer-generated voice. Just recreate it. And that's what I did. And there was no emotion involved; I just listened to the sound of the computer, and I repeated what I heard. But as the game continued, I was supposed to get more and more emotion into the computer, into GLaDOS. And she developed into a very sweet, passive-aggressive person.
— Ellen McLain: Voice of GLaDOS

#103

As an actor, you know, you come in, you do what you're told; You want to make your money. And sometimes you get good direction, and sometimes you don't get good direction. But it doesn't matter. And you certainly don't tell the production team any of this. But you come in and they'll, you know, they'll give you a lame line reading. And you're supposed to find some emotional value in that. But this is not what this team did. They would come in, for example, [and give me the direction] 'explosively indignant'. That's a wonderful direction for an actor. So, I must say that the production team won me over. Because so often I get bogus direction. But this time, I was actually directed and had fun at all the sessions.
— Ellen McLain: Voice of GLaDOS

#104

Sarcasm. GLaDOS is sarcastic a lot. And that's so much fun! But while I was trying to do this, trying to build in all these emotions, I still had to maintain the computer sound and maintain the pronunciation of repeated phrases. You know: 'Aperture Science' and 'The Enrichment Center'.
— Ellen McLain: Voice of GLaDOS

#105

I went to Portal's, umm... I Googled 'Portal in-game,' and I went to the trailer. And I just loved it! I loved hearing my voice! [Laughter] And as I told the production team, I have never played a computer game in my life. But I want to play this one! [Laughter]
— Ellen McLain: Voice of GLaDOS

#106

In the sound booth, you're only connected to the production team by an intercom. And they can talk about you, and you don't know what they're saying. But you trust that they're being kind. And, you know, they'll... They'll push a button and then they'll say something to me. They say a direction. 'No, that wasn't right. Do it again. Give us three more reads.' And a lot of time what happens is they'll... They would play the computer-generated sound for me on one line. And then I'm supposed to recreate that three times with the director saying, 'Ok, well, you know, more sarcastic,' or, 'More angry,' or, or, 'More irritated.' So: You stand in the booth hour after hour after hour. They're always very nice; they let you have water in here, and maybe a pencil to take a few notes. And of course you want it to be interesting, because you want them to hire you again!
— Ellen McLain: Voice of GLaDOS

#107

So when they told me that there was going to be a song at the end, I thought, well, allright, who's going to write the song? And they told me there was going to be a song written by Jonathan Coulton. And I listened to a song that Jonathan had written, and it was very funny, very clever. So, I thought at that point, well, you know, this'll be okay. But I am an opera singer, so usually I sing: [opera singing]. And I thought, well, will I be able to have the right style for the song? So I was concerned. But then before the recording, they sent me an mp3 file of the song. And I listened to it with Jonathan singing it. But I loved the little song. And at home, as I practiced the little song, I tried to, you know, get back to GLaDOS's voice. You know: 'Aperture Science...' Just this tiny little passive-aggressive computer who's all alone until people try to come in and murder her. So of course she gets upset! But she seems... She seems to have this real affinity for cake. And... And I want to play the game because I want to recreate the cake recipe. And then put 'Portal' on it, and be able to serve it to my friends when they come over to my house.
— Ellen McLain: Voice of GLaDOS

See also