Matt Gilbert (who blog-watchers will remember as the creator of the scrollbar scarf) has been thinking about speech synths: "Speech synthesis is usually about converting text to speech, but what if you approached it differently?" He designed a great-looking instrument out of curved wood, string, rubber bands and faders. As you press on the strings, it triggers a Max/MSP patch which emulates a voice. It's can't speak, but gets pretty close to singing. As ever, this makes no sense until you see this slightly freaky demo video [QT](thanks, Roberto)