17.2 Choosing Your Voice Actor

Careful selection of a voice actor is a necessary ingredient in producing a successful, branded user experience. The following are important considerations for choosing an actor who will be appropriate to your application.

17.2.1 Professionalism and Experience

The voice-over industry is vast, comprising areas as diverse as radio, animation, CD-ROMs, documentary films, industrial training videos, airport announcements, and informational videos, to name a few. VUIs, however, present unique and significant challenges. For example, when a voice actor records an instructional video on in-flight safety and emergency procedures, the meaning and context of each utterance are self-evident, and so most actors will not need much direction. In voicing a speech application, however, it is often not obvious how a given recording will fit into the larger interaction between the system and the user. The voice actor must make a conscious effort to internalize the context of each utterance, always thinking, "How will this recording sound in context?"

In addition, VUIs are notorious for long, tortuous help prompts that contain many disparate bits of information. The voice actor should be able to communicate such gnarly bundles of information to the user with an air of effortlessness. Voice actors who are experienced with speech systems are able to quickly grasp the meaning of these long prompts and deliver them with just the right inflections to guide the listener smoothly through the content.

The ability to manipulate prosodic variables, such as pitch and rhythm, is highly relevant to the success of concatenation-intensive VUIs. The voice actor should have an ear that not only can distinguish different intonation contours but can also produce them without sacrificing the desired energy, tone, mood, or persona. An experienced voice actor should also be able to perform stress shifts in isolation for example, "SEVenteen" versus "sevenTEEN."

Trained voice actors are sensitive to different kinds of variation in pronunciation. Having received proper direction, the trained actor should stay on phonetic course. The persona design may dictate pronunciation differences based on socio economic variation, regional variation, ethnic variation, or stylistic variation.

17.2.2 Coachability

Some musicians have a technical understanding of the music they play and of the instrument they play it on, but others do not, although they can still, say, "bang out a tune" on the piano. A similar dichotomy exists among voice actors. It is worth your while to find a voice actor who has technical knowledge and control of his or her instrument and who takes technical direction well, especially if your VUI relies on concatenation. A voice actor who has poor voice control, who cannot match an intonation contour, or who does not take direction well can turn a simple three- or four-hour recording session into an all-day affair, including the time it will take for extra editing on fumbled takes. Of course, you can find out how coachable an actor is by simply holding an audition (discussed in a moment).

17.2.3 Fit with Persona

Choose a voice actor who captures the desired persona for your application and who will be able to project that character consistently and effectively over the phone. An actor's suitability for a particular persona can be determined in a brief audition.

17.2.4 Demo Tapes (or CDs) and Auditions

Demo tapes and CDs are essential for narrowing the selection of voice, but they are no substitute for auditions. Auditioning a few key performance areas should provide an accurate picture of the actor's skill level and suitability for your speech application.

• Determine the actor's ability to interpret the persona by recording a few sample dialogs, including "best path" and error scenarios. In recording these persona samples, you may even find that the actor will bring greater depth to the character than you may have anticipated.

• Determine the actor's ability to take direction for concatenation-intensive messages, which include dates, times of day, money amounts, travel itineraries, driving directions, and so on.

• Whenever anyone reads a passage out loud, listeners can tell how well the reader understands the content. When someone, even a voice actor, reads a passage that he or she comprehends poorly or not at all, this too is obvious. It is therefore important to test the actor's ability to convey meaning in the persona under more difficult linguistic circumstances. Have the actor run through some longer informational messages, such as help prompts or a tutorial.

• Listen carefully for any impediments that will be noticeable over the telephone, excessive "mouth noise," "wobble," lack of vocal control, or any other vocal characteristics that might be received unfavorably.