Background music should sit under the speech and come back up when the talking stops. This is called ducking. Doing it by hand means four keyframes for every piece of speech; Premiere Pro can do it for you.
Last updated:
Add keyframes to the volume line on the music clip with the Pen tool (P) or by Cmd/Ctrl-clicking: one just before the speech, one just after, then pull the two points between them down. For short videos this is the most controlled way.
Premiere Pro can generate these keyframes itself.
If you cut or move the speech afterwards, the keyframes stay where they were; press Generate Keyframes again.
Auto music ducking in Studio › Audio does not ask you to tag clips one by one: it finds the speech on the other audio tracks itself.
For music under speech, a drop of 10 to 15 dB works in most videos. If the fade is too short the music sounds like it is pumping; too long and the start of the speech gets buried. A fade under half a second is a good starting point.
With one or two pieces of speech, manual keyframes are enough. In a long video, Essential Sound's Ducking or auto music ducking turns minutes into seconds. Whichever you use, listen to the result once from start to finish.
Captions and Graphics are free. macOS and Windows.