I've long applied a smidge of dynamic compression on just the loudest bits of my recordings — say, a 4:1 ratio on everything over -4dB, after normalization — to create more headroom to increase the volume of the overall recording without affecting the sound. (In my experience, most of those high peaks are things like stray extra-loud snare hits that don't sound much different if dynamically compressed a bit.) But I've also wanted to find ways to use compression to address another issue: shows where quiet songs are interspersed with loud songs, to the point where when listening back on speakers you're tempted to continually twiddle the volume knob so that you can hear everything at a pleasant level.
The results so far have been ... not great. If I apply a low compression ratio to everything above the peak level of quiet songs (say, -12dB), the levels work nicely on speaker playback: the quiet songs sound quiet, the loud songs sound loud, but both are nicely audible at living-room playback levels. On headphones, however, the loud songs, which got hit by the compression stick, begin to sound faintly unnatural, with less room to breathe than in the uncompressed versions. I'm finding that's the case even with very low compression ratios, by which point I'm not adding much volume to the quiet bits anyway.
Any suggestions from more experienced audio editors for tricks I could try, or is this just an inherent tradeoff I'm going to have to live with? I recently asked a professional audio engineer friend about this issue, and he replied that “the art of knowing how to use compression well is 90% of what I do, and can’t be replaced by a plugin,” which was simultaneously reassuring and not at all helpful…