Game Development Reference
In-Depth Information
Ideally, your tools and pipeline allow for narrative designers to drop new
VO into the game without the need to rely on any other team members,
such as programmers or audio experts. In most systems, these files will be
digitally generated, text-to-speech files (à la Stephen Hawking's artificial
voice).
Not all text-to-speech is created equal, however. Low-quality, excess-
ively robotic-sounding text-to-speech does a poor job of standing in for
human VO. It won't reveal important issues and will tend to be ignored by
nearly everyone on the team due to its grating, barely comprehensible
nature. Cheap-sounding text-to-speech will make any well-written line
sound bad and prevent accurate evaluation prior to recording. Worse, by
making every line sound clunky, it can unfairly hurt the credibility of the
game writing.
High-quality text-to-speech, however, can be a boon to the team's
many storytellers, and help them more accurately test the quality of the
dialogue writing and event-triggering decisions. If you're considering cut-
ting corners by not springing for the best digital voices available, think
again. It's probably more than worth the extra few dollars.
Repetition and Tuning
The noticeable repetition of a video game VO line can cause a barks sys-
tem to suddenly go from making the world feel real and alive to blatantly
revealing that same world to be artificial and contrived. There are few oc-
currences that can so immediately, decisively shatter a player's suspension
of disbelief, reminding him that he's playing a game.
Game writers and audio designers with experience in this area know
how important it is to write a sufficient number of variants of each type of
line so that repetition is reduced or eliminated. The key fact that can be
difficult to know early in the development cycle is:
How often is this trig-
ger event likely to happen?
The more often it occurs, the more variants for
the associated VO line will be needed.
However, the brute force method of simply writing more and more vari-
ants for triggers that keep calling a certain line is not a very efficient way of
dealing with the issue. Also, there is a point of diminishing returns in terms
of writing quality when trying to generate more than, say, twenty ways to
say the exact same thing. Integrating tools that allow for monitoring of
how often each line is being played in the game, and for tuning these
