Philips has an interesting project featuring (immobile) robotic cats that can display a wide array of facial expressions to be used as social cues in human-robot interactions. True, the cat may be sausage-lipped, tell bad jokes, and move in a slightly jerky fashion, but it's definitely a step in a very interesting direction, and I for one applaud them for taking it. (See the site for more info, images and movies to reach your own conclusions.)
I was somewhat surprised though that when they tested different "personalities", they didn't also run a separate series of tests to learn about the (human) test subjects' reaction to different voices — or if they did, this wasn't mentioned. I was distinctly taken aback by the cat's voice, feeling it not only didn't match the electric animal's appearance, but also finding it rather stern. While I realize that speech synthesis has its own issues, choosing an amiable voice to work from would go a long way towards acceptance by the more aurally oriented.