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

How long will the recording process take? It varies, depending on many factors. But a good director, using software that makes it easy to record a large number of prompts quickly, along with professional, experienced voice talents, can record about 200 prompts per hour . (This average includes every kind of prompt, from short numbers to long instructions.) That's important to consider, because time can be a major cost factor. As of this writing, many nonunion talents charge from $100 to $400 per hour, while union talents charge about $1,300 per hour ”not including studio time and casting costs. Most directors can assume that a mid- sized application (one designed to deliver flight information or a simple home banking application) requires about 14 hours of directing. The number of hours will vary a lot, depending on how good the voice talent is, how good the director is, and how many mistakes need to be corrected in subsequent recording sessions. So a budget between $3,000 and $5,000 is reasonable for nonunion talent, while a budget of $12,000 to $18,000 is more likely for union talent.

When recording voice talents, it's always better to make the recordings digitally in a high-quality format ”by recording them directly to a computer or other digital media ”even though the final product will be played over a low-bandwidth device (the phone). There are two reasons for this. First, it allows the recordings to be repackaged for other uses, such as TV and radio spots or presentations. Second, recording and saving the prompts in a high-quality format enables them to be down sampled (changed to another format of lesser quality) in many different ways for use by any speech-recognition system. This minimizes any loss of quality that may occur if, for example, a designer recorded the prompts in a low-fidelity proprietary format. Doing so would prevent the designer from getting the best quality when those audio files were then imported into a different software environment.

Finally, what happens if the recording session is completed ”and later it's discovered that the designer needs to record one or two minor prompts? This may necessitate setting up another studio session with the voice talent at a substantial minimum cost. However, many voice talents have their own home-recording systems ”and with today's digital technology, even a home system can deliver high-quality audio. In that case, the designer can simply ask the voice talent to record the missing prompts for a small fee and forward the files via e-mail.