Been mixing a bunch of videos of talking heads with a music track underneath (which is a really obnoxious and loud rock track) - the general structure is that we open on the music intro for ~10 sec, then go into several chunks of talking heads with small ~5-7 gaps in between the chunks, music underneath, and finally a ~10sec ringout on the music at the end when the people are done yakking.
Getting tired of keyframing the music level, I’ve been using sidechaining to control the Cubase compressor on the music track – but you can’t slow down the attack enough (and to a lesser degree the release) for my tastes to duck the music under the sync sound nicely (I’m feeding the sidechain a clone of the sync track that is slipped back half a second or so to make it duck in ahead of the actual sync sound track.) I’ve played with adding a short echo with a long decay or using a really hollow long reverb to feed the sidechain as well, in an attempt to use it like an ‘average’ function to keep the compressor from reacting too quickly while someone is talking, but it’s not quite working out as well as I’d like either.
I guess ideally I’d like some sort of way to get an RMS-type sampling with a fairly long average to be able to feed to the compressor or limiter sidechain on my music track, to ride over the small gaps in people’s words and sentences but still allow the music to ride up a bit during the longer graphics breaks between their chunks or at the end.
If anyone has any suggestions or a different way to approach this I’d appreciate hearing them 
Or should I just give up and be looking for a 3rd-party compressor that would give me much longer attacks and look-aheads?