Ducking

The technique is old and the automation is new. A sidechain compressor listens to the voice track and pulls the music down by a set amount whenever speech is present, with attack and release controls governing how fast the dip arrives and how gracefully it recovers. Most editing software now offers this as a one-click auto-duck. Set it too aggressively and the music pumps audibly under every word, which distracts more than loud music ever did. A dip of a few decibels with a slow release usually does the job, and if the music still fights the voice after that, the mix level was wrong to begin with.