Ben Gonshaw: Digital Media Theorist & Game Design Consultant
Home Latest Updates Design Rants Musings Links Comments

LA PLUIE CA CHANGE

5th January 2005

Generational Leap

The UK networks wheel out Singin’ in the Rain every Christmas and it never ceases to be magical. For those of you who have not seen it, it is a musical made in the fifties, that tracks a movie studio during the difficult transition from silent movies to those with sound. In the movie, set in the twenties, new technology comes along and wipes out many studios and destroys the careers of those stars who looked good at exaggerated miming but cannot act naturally and hold a conversation.

Movies without speech cannot accomplish very much at all, as Kathy Seldon says in her first appearance in the movie:

“ if you've seen one, you've seen them all...Oh, no offence. Movies are entertaining enough for the masses, but the personalities on the screen just don't impress me. I mean, they don't talk. They don't act. They just make a lot of dumb show...”

Playstation 3 or even 4 may not produce such a content revolution as the talkies did for movies, but the principal remains the same. A huge shake up has been going on in the industry, as the demands of next generation technology are being felt. However, there is more to the story than that. Forget that games are starting to look more realistic, that they have 7.1 sound and convincing physics, what people are waiting for is a generational leap in what games are. A whole heap of ‘dumb show’ is all that games can accomplish while they do not handle narrative and personalities in a convincing manner.

Getting There

The idea of games as a narrative journey is not a new one. First text adventures and then point and click graphic adventures tried to give the player a narrative thrill. Now cutscenes and scripted sequences inserted into action games have taken over. However, until the narrative is interactive and until the other personalities actually have personality, games will not advance significantly. For this to happen there are many barriers to overcome before virtual people and the stories they create can be convincing. Some are prevented by processing power while others are software problems that have not yet been sufficiently solved.

1: Global Stories

It is all well and good for the story to work if the player is shepherded along a set path, but the true revolution will come when the story unfolds dynamically based on the player’s actions. The player should take an active part in creating the series of events that end up forming a story. To do this, the game needs to be able to create a convincing and exciting story for them to take part in. This does not mean that the game needs to know what makes a great story, as it could be an emergent property of having many actors interacting. Forcing the global story through brute force is time consuming and often has unsatisfactory results.

2: Convincing Actions

This is an incredibly difficult task to accomplish. Even in the limited domain of shoot or die, enemies make basic, laughable mistakes. It looks stupid when you can repeatedly shoot someone in the foot from a distance and they never move, until they collapse dead, and even then their colleagues are still stood stock still. For a narrative to work, the actors in the game must make convincing decisions of an abstract nature, such as in Sims 2. These actions will create the story above. There is much active research on this topic, but the results are still far from convincing.

3: Natural Language

In point and click adventures characters would never get bored of talking to you. In fact, they would patiently and robotically repeat the same phrase over and over verbatim. Of course, this was absurd, but acceptable. However, when the actions are generated dynamically, and the possible situations cannot be predicted by the game’s creators, then convincing natural language becomes a necessity. It would be impossible to predict what characters would need to say before it happens, as we cannot predict all the possible actions that may occur in a single session, let alone in all sessions. Natural language is also a hard problem in AI, although research continues apace.

4: Natural Speech

So the character has had things happen to them and decided a course of action. The character has worked out how to communicate that to other actors and to the player. Now they need to be able to say the words and sound human. More than that, they need to be able to express emotion, to whisper or shout, in fact to act as well as Pacino or Spacey. You may be suprised to hear that this is a problem that has been sufficiently solved already. However, the computation needed for a single voice is far in excess of what is currently available in a console or PC, requiring super-computing levels of power. Right now we could not do a single word in real-time, let alone an argument between two people or a fight between hundreds.

And Finally

These are the issues that I would class as being the key stumbling blocks to the real games revolution. How long it will be until the hardware is up to it and the software is sophisticated enough could keep experts guessing for a long while. Until this happens, games makers will be using all the dry ice and as many mirrors as they can lay their hands on to fool people into thinking that they can do it. For now we will have to make do with advances such as real-time, physics based animation and more convincing lip syncing.

In Singin’ in the Rain, Lena Lamont is the pre-talkie movie star of choice. She does not have a speaking voice that can be used in the new talkies, and she cannot act any more subtly than the Stay Puft marshmallow man. In the movie she lost her position as the queen of the screen because of her lack of substance. If she was a character in a game, her career would be safe for a long while yet.

Write Comment : Back to Musings