I've been making steady "progress" on my colorstream sound visualization project. Its algorithm for colorizing speech now has two parallel strips. A larger, upper strip represents the "whispering" range. (Even in a whisper, the ear can distinguish most sounds & words, so there's lots of data in there.) And a thinner, lower strip represents the vocal range - a rainbow, from the lowest bass note to the highest soprano note. (With this strip, one could tell apart "z" and "s" sounds for example, as one is voiced and the other is not.)
Here are some example movie clips I've been experimenting with.
The opening sounds & narration from the Transformers movie (2007) from this video can be seen colorized here. [Update: source video has been taken off-line. Audio-only is cached here, for purposes of this demo.]
And here's a colorized view of the "You can't handle the Truth!" scene from A Few Good Men. Enjoy!
Here is something of a phonetic color key for how different sounds match up with colors.