Like speech input, speech output is a familiar and natural form of communication, so it is also an appropriate complement in a character-based interface. However, speech output also has its liabilities. In some environments, speech output may not be preferred or audible. In addition, by itself, speech is invisible and has no persistent output, relying heavily on short-term memory. These factors limit its capacity and speed for processing large amounts of information. Similarly, speech output can also disrupt user input, particularly when speech is the input method. Speech engines generally have little support that enables the user to interrupt when speech or other audio has the output channel.
As a result, avoid using speech as the exclusive form of output. However, because Microsoft Agent presents characters as a part of the Windows interface, it provides several advantages over speech-only environments. Characters can be combined with other forms of input and output to make options and actions visible, enabling a more effective interface than one that is exclusively visual or speech-based.
In addition, to make speech output more visible, Microsoft Agent includes the option of authoring a character with a cartoon-style word balloon. Other settings enable you to determine how text appears in the balloon and when the balloon is removed. You can also determine what font to use. Although you can set a character's word balloon attributes, be aware that the user can override these settings.