Michael De Plater is the Creative Director of the upcoming Tom Clancy’s EndWar game. Three Speech writer Keri Allan caught up with him on a recent trip to the UK to discuss the introduction of speech recognition technology into the console strategy genre arena, and how this technology could revolutionise how we play games…

TS: So how did the idea of the voice recognition in the game come about?

MDP: Basically no-one has successfully done strategy on a console before: it’s like the last big genre that hasn’t moved from PC to console. So we knew we had to do something different on the interface, and also we wanted to make the experience different for the player. In essence we wanted to make you, as the gamer, the General.

Generally there’s been a big trend in games towards more natural inputs, whether it’s motion sensors, steering wheels with driving games and instruments with music games – like the guitar in Guitar Hero. We’re trying to let you play the role of military commander, so if you look at what an actual military commander does, its speech!

So we began to look into it, and started by looking at a game called SOCOM that came out about five years ago. It was a tactical shooter that had a level of voice command that worked pretty well.

We assumed that the technology has got more powerful over the last five years, as had the new consoles, so we decided to take voice control to the next level, solving the UI (user interface) issues whilst making strategy as a genre, work on consoles.

TS: So how tough was the development process and integrating speech into the game’s control system? There’s been many ‘blips’ with voice recognition technology in the past, did you have to face any issues of your own?

MDP: Well, we spent a lot of time on it. The hardest thing wasn’t actually the voice recognition, it was making the AI work with it. Basically, to make a game work with voice recognition it has to have a relatively small set of quite intuitive commands to recognise. So we had to make a relatively small subset of orders, then make the soldiers smart enough to follow those orders. You have to be able to give quite a high level order like ‘Unit 3, secure Alpha’, and then you have to trust that they’re going to go off and do that order in a fairly intelligent way.

TS: So how did it go working out all the commands what they were going to be? Did you think, how am I going to actually squish it down into small set of commands, is that quite straightforward?

MDP: We had to squish all sorts of things like the number of unit types, the number of units themselves etc. A strategy game has a lot of features. There is resource management, the victory objectives on the map, the navigation points, the buildings, the upgrades, and the resources themselves. We had to sit and work out the ‘tech tree’ for the game like: ‘Unit one, upgrade Alpha, air support’ etc. So we had to look to how and where we could streamline the controls by combining systems and commands.

It was really good though, it’s like Haiku! By compelling you to be more economical about it, I feel that we’ve actually ended up with a much better game.

TS: What about your choice of words and issues with dialect? Did you not have to deal with any problems in that area?

MDP: Well we worked with a company called Phonics who provided this middleware. Using this, we had to work on lots of optimisations and testing to get exactly the right vocabulary commands in place. For example, with the word target, we tested ‘target’, ‘reticule’, ‘crosshair’, ‘cursor’, to see which got the best level of recognition.

With the uplinks/bases like ‘Alpha’, ‘Bravo’, ‘Delta’, we had a maximum of 12 uplinks needed at one time, so out of all the 26 letters of the alphabet we chose the twelve that had the best recognition results. The only place you can’t do that in is the unit numbers. Obviously you have to use the ones that are given to you, but also they’re such common words that the speech system was pretty good at picking them up anyway.

TS: With the focus being on voice controls, did you find that you had to have a different mindset from the past when you began development?

MDP: I think that generally that’s something that’s true of console over PC anyway. On PC, every single time you do a new game the temptation is to change from the last game by adding more features. If you’ve got 100 buttons on the keyboard, we’ll just add another few combinations to offer more control options. You can’t do that on the console, you’re constrained by the pad.

Using voice is much better anyway I’ve found, because I think it forces you to cut things down to what are the interesting decisions and the fun things to do. You discover you can simply get rid of everything else!

TS: So now you’ve shown us that voice-controlling games is possible – and also fun - do you think we might see this kind of technology appear in more titles?

MDP: Speech is really a natural form of human interaction, and now that it can work quite well in a game I think we will use it more. I can see its use in sports games, and casting spells in an RPG would be another obvious use. It doesn’t solve everything, but there’s certain things you do in a game where using speech recognition is definitely a natural option.