The Art and Business of Speech Recognition: Creating the Noble Voice

Casting the right voice is one of the most important tasks in creating a successful speech-recognition system. The right voice talent must be appropriate for the application and reflect the overall brand of the client. A large bank might want to cast a voice talent for their advertisements that possesses a strong, masculine quality such as Charlton Heston. However, his voice may not be appropriate for their speech applications because it might sound too rough or intense . The right voice may have similar hallmarks to Heston's, but might sound more approachable and friendly. (See the section "Choosing the Voice of the Brand" later in this chapter for a more detailed discussion of this topic.)

One way to start the casting process is to call local recording studios or casting companies and ask for a demo tape or CD of several voice talents. It is reasonable to ask the talents to read a script that you have prepared so that you can hear the voice talents saying the exact text that would be used for a system. Listening to these samples will help narrow the pool of candidates. But listening to an audio sample will not reveal what it's like to work with a particular voice talent ”or how that person will sound reading the actual audio prompt script. So it's usually a good idea to audition the voices personally .

Typically, it only takes about 30 minutes to audition a prospective voice talent, having them

• Understand what they're reading

• Understand the context in which it will be heard

• Read several types of prompts that are representative of various parts of an application

For the audition, you should prepare a short script that includes either actual system prompts or ”if casting is done earlier in the design process ”made-up prompts similar to the ones that will be in the completed system. These can include basic prompts, such as:

"Welcome to Cool Startup, Incorporated." (To be spoken warmly)

"Please enter or say your PIN number." (To be spoken casually)

"I'm sorry, but this account has been temporarily disabled." (To be spoken sincerely)

It's also a good idea to include prompts that follow each other in a list, such as

"Thirty."

"Forty."

"Fifty."

and so on.

By having the voice talent read several prompts like these sequentially, you can determine whether the talent (either intuitively or with direction) can speak each number using the same tone and frequency, or pitch. That's important to know, because if the pitch of the talent's voice rises at the end of each word during the actual recording ”as if they were reading a list to someone ”it will make it harder for the recording engineer to concatenate and rearrange the numbers later. It's best to have a voice talent who can keep his or her volume, tone, and pitch consistent on prompts such as these.

You should also take some time during the audition to direct each voice talent and monitor their reactions . This can help you gauge how easy ”or difficult ”it would be to work with that person during a potentially long, painstaking, and repetitive recording session.

Cost is a key factor to consider while casting for voice talent. Most voice talents ”both union and nonunion ”charge by the hour . Union talent is generally more expensive, and their contracts can include stipulations and restrictions about working conditions, such as overtime charges, travel expenses, and so on. However, membership in an actor's union usually implies a higher level of professionalism and experience, both of which help to ensure a fast and successful recording session. Nonunion talent is generally ”but not always ”less expensive, because they set their own rates. And nonunion talent can be just as good as union talent. However, be sure that the voice talents have the skills and experience necessary to do this exacting work. In either case, a local recording studio or casting company can provide a more complete understanding about pricing for each.

What about amateur talent? Well, there are occasionally natural talents who sound just perfect and can learn on the job, but such people are few and far between. And as we're about to see, there's more to being a successful voice talent than just having a nice voice and being able to read.