Personality

VUI Personality

Every VUI has a personality, whether you want it to have or not. Having said that, it’s not all that easy to describe here in words, something that is assimilated acoustically. Suffice to say, people can read a lot from a voice - they can judge the age, the mood, the confidence, and the patience of the person. Callers pretty quickly create a mental image of the person behind the voice - it could be a middle-aged efficient business assistent, an over-excited young girl, an impatient sligtly arrogant man. It's not just the words that matter - the voice that you choose to deliver those words will have a major role in the caller's perception of the service.

As I said above, what makes a voice good or bad is very hard to describe - and I don't have the wherewithall to put any examples up on the web-site. Well, not without infringing copyright and upsetting some "voice talents". However, here's one or two experiences that we've had. One company had a preferred speaker - he did all the advertising voice-overs for the company. However, when it came to VoiceXML applications, his tone was over-bearing and basically too aggressive. So after much toing and froing, we convinced the company to move to a more soothing female voice. There were no customer complaints. Another female voice that we used was slightly too impatient - nothing too bad, but there was a slight edge to it which could be a little annoying after a while.

Directing "Voice Talent"

Most recording studios have a list of voice actors, who quite often specialise in advertising or documentaty voice-overs. They should also have a sample CD to help you choose which voice best suits your application. If you're working with a large company, they may have a fixed company voice that you have to work with - this isn't always a good thing! If possible, try and get several speakers to record some of your prompts, so that you can judge how well they suit your application.

It's a very good idea to go to the first few sessions with an actor, as they are often unfamiliar with the requirements for a VoiceXML application - in particular snippets and the necessity for open and closed versions of many snippets. So quite often you'll find yourself explaining the whole application to the actor, as well as the motivation behind the wording of the prompts and expected caller profile. Good actors have many different voices - ranging to from peppy enthusiastic advertising voices to serious slow voices. Giving the actor enough back-ground information helps them choose which voice is best suited to the application.

Unless you're actor yourself, it can be a little daunting directing a real actor. However, as the designer of the application, you have the best idea of how you expect the application to sound - it's your creation after all. Expressing what you want is often very difficult - particularly if you are not familiar with the terms that an actor would use for various qualities in a voice. Very often you may well be reduced to trying to say the prompt yourself the way you want it - and hoping that the actor can gleam enough from your "performance" to produce what you want! It's helpful if there is some frame of reference, such as a well-known voice from TV or films, that you can use as an example. You may need to be quite resolute about achieving the sound that you want - although most actors understand that there is more than one way to say a prompt. So you'll need to be clear about where you want the emphasis to be - as well as how fast or slow the prompt should be - whether you're looking for an enthusiastic voice, a supportive one or a serious one. It's your choice!

It's also helpful to talk to the sound engineer and make sure he's clear about what formats to deliver in and what pre- or post-processing is necessary (recording prompts).

If you have any comments, ideas, issues, etc. about this topic why not try the voice-push forums