The idea itself is pretty good. And it might even be fun to make. You'll be needing a database with sounds, that much is sure. I wonder though, is your output one mp3 with all kinds of sounds mixed together, or are you offering a complete environment with variable inputs, so I can add that to my engine?