When reverb masks the next line

The plate that sounds great in solo

You put a long plate on the lead vocal and solo it. It sounds expensive: every phrase opens into a soft cloud that hangs in the air. Then you bring the band back for the chorus, and the first words of every line are hard to catch. The singer did not get worse. The tail of the line before is still sounding when the next one starts.

Why it matters

The tail is the same voice in the same band

A reverb tail is mostly the last word, smeared over time. It has much the same spectrum as the voice, so it competes with the next words in exactly the bands they need. In solo nothing else is playing, and the overlap sounds like space. In the mix the tail adds to the guitars and keys that already cover those bands, and the old tail is at its loudest during the first words of the new line, the ones the listener needs to catch.

Shortening the decay fixes the overlap, and the lesson on reverb and emotional distance shows how to fit it to the tempo. But a shorter decay also shrinks the bloom in the gaps, which is the part you liked. Ducking the return and replacing part of the reverb with a delay let you keep that bloom and move it to where no one is singing.

The science

How much is left when the next line starts

The next words are masked by whatever shares their bands at the same moment, and the louder the masker in a band, the louder a sound must be to stay audible there (Fastl and Zwicker, 2007). So the question is how loud the old tail still is when the new line begins.

A reverb tail falls 60 dB in its decay time $T$. After a gap of $g$ seconds it has fallen:

With 0.4 s between lines, a 1 s reverb has fallen 24 dB, but a 3 s reverb has fallen only 8 dB. If the tail starts 10 dB under the dry vocal and the next line is sung at the same level, the old tail sits only about 18 dB under the new vowels. The consonants that carry the words are shorter and quieter than the vowels, so their margin is smaller still.

A compressor on the reverb return, keyed from the dry vocal, adds its gain reduction to that margin, and it does so only while the singer sings. At the start of the new line the dry vocal triggers the compressor and pulls the old tail down. When the line ends the compressor lets go, and the tail rises in the gap with its full decay. Eight dB of ducking turns the 3 s reverb's 8 dB of fade into 16 dB. The same sidechain idea works on many other pairs, as the lesson on sidechain beyond kick and bass shows.

A delay works differently. It makes discrete copies, and you choose when they land. With a feedback gain $k$, each repeat is $20 \log_{10} k$ dB relative to the one before (Zölzer, 2011): feedback at 0.35 gives about -9 dB per repeat. Each copy also repeats the words, and Brungart (2001) found that a competing phrase interfered most with understanding when it was spoken by the same talker as the target. A clear repeat of the last line under the new line is a quieter version of the same thing: the same voice saying different words. That is why I use a delay mostly as a throw: sent only the last word, timed so the repeats fall in the gap.

In the demo, switch between the plain reverb, the ducked reverb and the delay throw, and listen each time to the first word of the next line. The throw sends only the last syllable of each line to a filtered tempo delay. Try it at a dotted eighth, then at a quarter note, where the second repeat lands on the next line's first word.