The Use of Voice in Portal 2

By Kenneth Young

Portal 2 represents a real milestone for me. It’s the first game that, prior to even getting my grubby mitts on it, had me more excited about its prospective use of voice than any other aspect of the project. It’s a Valve game, so you trust that it’s going to deliver engaging gameplay and you know that the use of voice is going to be considered and interesting because, time and again, you’ve seen Valve push themselves and push their medium. But it’s also a sequel, and this lowers your expectations, especially as it’s the sequel to something quite so wonderful as the original Portal. The core gameplay couldn’t be reinvented, and Valve were clever enough not to mess with it; they added some new bells and whistles to keep things interesting in the story mode, but the most significant new draw of the gameplay in Portal 2 is its multiplayer which is segregated in to a separate experience.

Portal 2’s story, however, must have presented a bigger challenge. It couldn’t just be more of the same because we’d already had that experience. But the magic formula they’d discovered was so darn potent there was no way they could escape from it. Story and gameplay in Portal were intertwined in such blissful harmony; the repetitive nature of the puzzles and formulaic interactions with GLaDOS, their masterful development and gradual disintegration, culminating with the disintegration of GLaDOS herself, is what made the game tick and boom. So, faced with similar gameplay but unable to simply “do a GLaDOS”, Valve made Portal 2’s story a deft exploration of different characterisations which all fit over the same core gameplay.

The vast majority of articulate videogame characters are perceived primarily through their voices, and the characters in the Portal universe appear to be no different at first glance. But whilst your average videogame character suffers from the affliction of being an intelligent voice bolted on to an idiot meetbag puppet (which all too often isn’t so much an example of “The Uncanny Valley” so much as it is just plain “broken”), Portal makes full use of the power afforded by the disembodied voice.

I’m sure if you were to ask a fan of Portal to describe what GLaDOS looks like they would attempt to articulate what you see in the image above. But that’s the least of GLaDOS – that shell really only exists to give you something to focus on and destroy at the end of the first game. And that shell doesn’t even belong exclusively to her, it’s simply the mechanism through which her AI is connected to the Aperture Science facility. When GLaDOS is in charge, she becomes the voice of the facility, and the facility her body – you can’t separate the two from each other. When you enter a test chamber and see a barrage of wall panels rotating in to place – that’s her. When those wall panels stutter and malfunction it’s a sign that not everything is quite so clinically precise as it may have first appeared – a chink in her armour perhaps? And that camera on the wall…

GLaDOS is a brilliant homage to HAL, the sentient, murderous AI in 2001: A Space Odyssey. If you were to describe HAL as “a red camera lens”, you’d kinda be missing the point – he is first and foremost a disembodied voice and that is what gives his character all of its power. HAL is all-seeing and all-knowing; a true acousmêtre.

Cave Johnson, the founder of Aperture Science, is the most disembodied of all the voices in the Portal universe. His voice isn’t that of a sentient AI and he certainly can’t reveal himself Wizard of Oz style, for his is a voice from the grave. Your whereabouts within the proto-Aperture Science labs trigger instructions from, and the musings of, Johnson, the old voice recordings being contemporary with the dilapidated, vintage surroundings. Perhaps the most interesting aspect of this setup to consider is that despite the story playing up the primitive nature of this technology (in order to contrast it against the sophistication of the modern Aperture Science facility) it is in fact identical to that used in the real world by Portal 2 (and all other videogames); the simple triggering of pre-recorded voice samples as the player moves around, tripping switches. Considering that the in-game implementation is essentially identical for Cave Johnson and GLaDOS, it’s testament to the writing and design that only Johnson is perceived to be a mere voice recording.

“Hello? Are you there”?

Gun turrets are the only consistently embodied voices to be found in Portal and Portal 2. It’s interesting, then, that they are written to be so overtly stupid. They certainly aren’t intended to be perceived as complex AI’s, they’re simply laser sensors with guns. But the point is that embodied voices in videogames are a lot harder to sell as intelligent beings because the body they inhabit is inevitably going to be relatively stupid. For example, compare the super-intelligence of GLaDOS to the annoying tendency in other games of AI ground troops to get in your way despite them being a far more sophisticated piece of game technology. It’s no coincidence that GLaDOS is perceived as having such a sophisticated personality despite lacking a physical body – it is precisely because of the disembodied nature of her voice that the writers are able to pull this feat off quite so convincingly (assisted by lots of linear, scripted sequences, naturally). The use of voice on the gun turrets is clever because it doesn’t try and make them something they’re not – it acknowledges their intrinsic lack of intelligence, something the player would soon uncover during gameplay, and makes their character stronger as a result.

The multiplayer characters in Portal 2 aren’t a million miles away from the gun turrets – they are also overtly dumb, in a vaudeville comedy double-act fashion, but perhaps even more so due to the fact that they communicate via robotic vocalisations rather than speech. Their most explicit communication happens during cutscenes, seen from the voyeuristic point of view of GLaDOS’ spy-cams. It’s interesting that Valve allowed themselves to take control away from the player here, both through the use of these cutscenes and through relinquishing their love of the silent protagonist, but these seem like compromises which are necessary in order to create an engaging two-player experience. I’m sure the decision to use robot gibberish was primarily driven by not wanting to alienate the player from their character.

And then there’s Wheatley. Once Wheatley takes over the facility in Portal 2 he’s effectively just a stupid version of GLaDOS, which is certainly a fun juxtaposition (hilarity ensues etc.), but it’s essentially an exploration of the “turn everything on its head” approach that is the stuff of many sequels. No, the most interesting use of voice with Wheatley’s character comes at the beginning of the game prior to this transformation.

Initially, when you first meet him, Wheatley is not a disembodied voice, he’s a well-meaning little robot AI chappy; a discarded Personality Core, literally an earlier, inferior version of GLaDOS. He doesn’t have the same monotonous robot voice associated with GLaDOS or the gun turrets (or any of the other personality cores encountered in Portal 2 for that matter); this instantly sets him apart as being more likeable, more human, reinforced by the fact that he’s trying to save your life, albeit rather incompetently. The nice thing about his characterisation is that it brilliantly and immediately addresses the problem of the player as a silent protagonist; Wheatley has verbal diarrhea. He doesn’t shut up. He can’t shut up. Every potential awkward silence is filled with utter nonsense by this driveling idiot. It’s a genius idea, and a great performance by Stephen Merchant. The only problem is that by setting this precedent, when you eventually pick Wheatley up and carry him around, it’s a bit awkward to have him in the middle of the screen, staring at you in silence – it would have been better to make his default position be staring ahead (which he actually does a couple of times), only having him turn around to speak to you. It’s hardly a big deal, but it stood out to me in stark contrast to the otherwise brilliant presentation and considered use of voice in the game.

Once Wheatley becomes omniscient, he comes and goes as GLaDOS did in Portal 1 – the same roll explored through a different personality. I found it a lot more interesting to see GLaDOS adopt Wheatley’s previous incarnation; a voice forced into a physical manifestation so that you can carry it about with you (as a potato-powered microchip stuck on the end of your portal gun). This could have been awkward, but by making her small and unobtrusive and, crucially, limited by her vegetable matter power supply, GLaDOS’ voice is also empowered to come and go as the puzzle gameplay allows. This might sound like an insignificant point, but bear in mind that the thing that permitted GLaDOS’ intermittent communication in the first place was the disembodied nature of her voice – to have found a solution (i.e. an excuse) which allows this to continue despite her altered state, and persistence in the player’s field of view, shows an attention to detail and respect for their player experience that most developers fail to give the attention it deserves.

It’s easy to overlook why Valve settled on these solutions in the first place – the characters don’t behave this way because they were thought up in ignorance of the game and then crow-bared into it, their behaviour was dictated by the requirements of the gameplay. They consistently follow two very simple rules – everything else is the result of a problem solving exercise (i.e. a design process) that endeavours to stay true to these fundamental tenets:

  • Don’t use voice to communicate information or story to the player unless they are able to listen
  • Keep the player engaged; don’t lose them

That first rule sounds pretty obvious but in practice it’s actually rather hard to abide by, especially if you’re ignoring the fact that you’re working on a game and instead pretending that you’re working on a film-like experience with an attentive audience. The most common faux pas is to have a character talk to the player whilst their mind is occupied with another task, e.g. during gameplay (you know, that thing that people do when they’re playing a game?). There’s a real conflict here, especially in games with meat-bag characters that follow you about – if you’re with another character then it’s awkward not to have them say anything (because this silence highlights the fact that they’re just digital meat-bags rather than the “real people” the designers want them to be perceived as). Unfortunately, as the amount of meaningless dialogue that is injected in to the game increases (in a desperate attempt to make the characters “come to life”), so does the player’s apathy for any speech they might hear irrespective of how important the information it conveys might be.

The solution regularly used to tackle this problem is to dovetail the gameplay with non-interactive sequences (i.e. cutscenes). The problem with cutscenes is that they neuter the player’s agency, which is one of the reasons why players can quickly loose interest in them – they’d much rather be playing the game they were just enjoying thankyouverymuch. In other words, many games’ solution to rule one is to go ahead and break rule two. Doh. Whilst cutscenes aren’t intrinsically destined to alienate players, they have a real propensity to go on for too long and throw far too much information at the player, a problem which is exacerbated by not finding elegant ways to give the player any information or story exposition during gameplay.

Valve are totally ninja at making the player feel like they’re in control when they’re actually taking part in an interactive cutscene. Much of the time you aren’t even aware of it; all those short walks at the end of each Portal test chamber are cut and dried linear, pseudo-interactive cutscenes. The task of moving from A to B without any obstruction isn’t difficult enough to prevent the player from being able to listen to what GLaDOS, or Wheatley, has to say. In fact, having been challenged to complete a puzzle, the player is smugly and actively looking forward to hear what their adversary has to say about it and this gives the writers a few seconds to smuggle in a little more character development or exposition as well as the laying down of the next gauntlet. It’s not so epically long that players will lose interest, and it certainly helps that the script is so engaging. These frequent, positive interactions buy the player’s trust and permit more extensive interactive cutscenes. It’s a beautiful, beautiful thing when the right balance is struck between gameplay and story, and both Portal games do this with aplomb.

The setting and characterisation in the original Portal had such a pure and convenient relationship with its gameplay that I’m sure it was easy for many to dismiss it as being irrelevant to their own work. But when such low hanging fruit is only just being utilised by our medium it’s pretty clear there’s lots of gold left to be mined in them hills. And as if to prove the point, Portal 2 invents yet more characterisations, styles of performance, writing and design that are perfectly suited to our medium and more or less unexplored by anyone else. Which is exciting to behold, but it’d be nice to see even a few competent copycats, if not a few more pioneers.

Portal 2 is a real milestone. It inspires me; I hope it inspires you too.

About Kenneth Young 21 Articles
Kenny is a multi-award winning freelance sound designer, composer, audio director and music supervisor best known for his work on Sony's Tearaway, LittleBigPlanet and Astro Bot games. You can read more of his writing here on or learn more about his work via his professional website,


  1. Long awaited blog update. Loved the game, it's dialogue and writing. I hadn't thought of that concept of Johnson being representational of how we trigger voice in games and how it seems rather primitive. They (the wizards) say film became a dominant medium of the 20th century when it learnt to speak, and that games will be the dominant medium of the 21st century when it learns to listen (or observe mute protagonists). So it would be interesting to see game dialogue down the road be more reminiscent of GLaDOS than Johnson. Valve are one of few developers that I find to be exploring the limitations of their medium and exploiting them for superior game experiences such as characterising AI as being dumb. The developer's commentary specifically at the start of the first Portal shows you how much they understand storytelling tools without feeling they need to be a film. I feel the real genius lies here and not in the "look at me" gimmicks of Heavy Rain for instance. One bit I also liked as simple as it may seem was the real-time use of attenuation as you bounced up and down from a spring to just catch parts of Wheatley's dialogue who is above you before you fall back down, missing the important part of hi sentences.

Leave a Reply

Your email address will not be published.


This site uses Akismet to reduce spam. Learn how your comment data is processed.