Why not simply build a module around this principle: http://en.wikipedia.org/wiki/Precision_rectifier#Peak_detector
with a 4053 as the switch, controlled by H-Sync and V-Sync and optionally a third trigger input.
wlam!
perfectly synchronized audio envelope…
Edit:
Ok, so after a little more thinking I can see why this is not the way to go: We’d rather integrate the audio signal over time than just look for peaks in it. Also, we’d not want to see the actual buildup on the capacitor but synchronized samples, so there should be a sample-hold circuit on the output.
Forget about it – but we should really build that sample-hold module…