Even though the auto-start audio has been working well for me, I'm going to be reverting to click-to-play recordings (because that is what I would prefer) on all of my auctions. I don't like the animated characters either... pretty spooky.
The main reason that I don't just record my voice and use HTML is because it slows loading time quite a bit (especially for dial-up connections). The service I use has Flash, so I don't have to worry about that. Also, the recordings take up a lot of bandwidth and Sellers Voice will host all of these recordings for me.