Apply audio ducking for speech

We’ve got an episode where the host speaks over music. Apply audio ducking so the music drops under their speech in the full 03:12 episode, prioritizing clear voice intelligibility for the producer. Keep music return level gentle, not fully muted, and run a quick playback check from 00:01:05 to 00:01:30 to confirm no words are lost.

Before I deliver this to marketing, make the ducking easy to approve: increase the host’s vocal presence by bringing the vocal gain up two dB, set the music duck depth so SNR is at least 12 dB during speech, and show a before-and-after clip of the worst-sounding section at 00:02:10. Point out any places where overlapping music stems prevent smooth ducking.

I applied auto-ducking and now during a chorus the music still punches through and clashes with the host. The creative director will call this out. I’m not sure if the problem is envelope settings or the music stem level. Do I lower the music stem, tighten the duck attack/release, or change the duck depth first? Which fix is fastest and least likely to create pumping artifacts?

Every project I spend extra hours re-tweaking ducking after reviews because reviewers complain about inconsistent levels across episodes. I want one change to my workflow to cut rework and make initial ducks passable for most stakeholders. What practical default settings and a short QA routine should I adopt so first-pass ducking rarely needs revision?