Less than the solo button suggests. My default chain on a pop lead is one plate between 1.2 and 1.8 seconds with 20 to 80 ms of pre-delay, the send filtered down to a band from roughly 250 Hz to 8 kHz, a tempo-synced delay doing most of the sustain, and the reverb return ducked about 2 dB whenever the dry vocal is speaking. Set that way, the vocal reads close and expensive at the same time, and the reverb only becomes audible as an object when the phrase ends. I mixed and mastered Stefania for Kalush Orchestra, and the vocal space on a record that has to survive a stadium PA and a phone speaker in the same week is built from exactly these four decisions.
Why does reverb bury a vocal before it flatters it?
Because a reverb tail is spectrally the same material as the voice that feeds it, delayed and smeared. The energy sits in the same 1 to 4 kHz region that consonants live in, so every dB of tail is a dB of masking aimed at the one element the mix exists to deliver. Worse, the tail fills the gaps between words. A dry vocal is mostly silence; those silences are where the ear resets and where the next consonant gets its contrast, and a long tail paves them over. Paul White's piece on creating a sense of depth in a mix makes the underlying point: perspective is built from contrast, and a vocal drowned in its own tail has none left. In practice the return ends up 3 to 4 dB lower than solo listening suggests, which is why every level decision on it happens with the full mix playing.
What does pre-delay actually buy?
Separation between the word and the room. With 0 ms of pre-delay the tail starts inside the consonant and the voice moves backward in the image; with 20 to 80 ms the consonant lands dry, the ear locks localization to the direct sound, and the tail arrives as decoration. The precedence effect explains the window: arrivals within roughly 2 to 50 ms fuse with the first wavefront, which keeps control of where the sound seems to come from, and for music the fusion stretches toward 100 ms. Pre-delay parks the reverb inside that zone on purpose.
I set it from the tempo. At 120 BPM a 16th note is 125 ms and a 32nd is 62.5 ms, and a pre-delay snapped to the 32nd makes the tail breathe with the track instead of against it. Ballads take the long end: 60 to 80 ms with a 2 second plate reads as a singer standing in front of the room rather than inside it. Up-tempo songs take 20 to 30 ms, because past that the gap itself becomes audible as a flam on the backbeat phrases.
When does a delay do the job better than a reverb?
Whenever the arrangement is already crowded. A delay occupies time instead of spectrum: a repeat at the quarter note with two or three feedback passes adds length to the phrase while leaving the space between repeats clean, which is why dense choruses carry a louder delay than reverb without losing a word. The oldest version of the trick is still the cheapest. Slapback echo, a single repeat at 60 to 250 ms with no feedback, has been thickening vocals since Sam Phillips wired two tape machines together in 1954, and a 90 to 120 ms slap at 6 to 9 dB under the dry vocal still does what a short room does, minus the midrange smear.
My working split: the slap or an eighth-note delay is on the vocal all song as part of its sound, the plate is automated, and a quarter-note delay with filtered repeats gets pushed 1 to 2 dB in the last chorus where the arrangement can afford it. The repeats are filtered harder than the reverb, down to roughly 400 Hz to 4 kHz, so they read as echo, never as a second singer.
How do you treat the reverb return?
Like a backing vocal that has to stay out of the lead's way. The send gets a de-esser first, before the reverb, because a plate turns every s into a cymbal splash; pulling 3 to 4 dB of sibilance on the way in costs nothing audible and cleans the whole tail. The return gets a high-pass around 250 to 300 Hz, a low-pass between 6 and 8 kHz, and, when the tail still argues with the voice, a broad 2 dB dip around 2 to 4 kHz, which moves the reverb behind the consonants instead of on top of them. The band-limiting idea is as old as plates themselves: the EMT 140 that defined vocal reverb from the late 1950s onward never pretended to be a full-bandwidth room, which is a large part of why plates sit so well under voices, and the design thinking behind that family of boxes fills Valhalla DSP's learning library, the best free reading on reverb design I know. Last check is mono. Fold the mix down: a return built from width alone disappears, and a tail that doubles in apparent level in mono is phase trouble a stereo room was hiding. The de-essing half of this chain has its own article in how I de-ess without dulling the voice.
How does the tail stay out of the words?
It gets ducked. A compressor on the reverb return, keyed from the dry vocal: fast attack, release between 250 and 400 ms, threshold set for 2 to 3 dB of gain reduction while the singer is speaking. The tail tucks itself under every phrase and blooms in the gaps, which is what engineers used to ride by hand on the echo return, and the reason old records sound dry and huge at once. Capitol did a version of it with real rooms under the sidewalk, covered in the echo chambers under Hollywood and Vine. Automation finishes the job: the send drops 1 to 2 dB in verses, rides up into the final chorus, and gets one deliberate throw, a single word sent hard into the delay at the end of a phrase.
What are the starting points by song type?
A pop lead at 100 to 125 BPM: plate at 1.4 seconds, pre-delay on the 32nd note, return filtered 250 Hz to 8 kHz, eighth-note delay underneath, duck at 2 dB. A ballad: long plate at 2.0 to 2.4 seconds, 60 to 80 ms pre-delay, quarter-note delay carrying most of the chorus. An up-tempo dance record: slap at 90 ms plus a quarter-note delay, plate under 1.2 seconds and 4 dB lower than instinct asks, because the arrangement is the reverb. These are where the fader starts before the mix votes. The chain in front of all of it is in the vocal chain in order.
Frequently asked
What pre-delay should a lead vocal reverb have? Between 20 and 80 milliseconds. Short enough that the tail still belongs to the voice, long enough that the consonants land dry before the reverb starts. Tempo is a good guide; a 32nd note at 120 BPM is 62.5 ms.
Should a lead vocal use reverb or delay? Both, doing different jobs. The delay carries length and excitement because it lives in time, not in spectrum, so it masks less. The reverb carries the sense of a room. Up-tempo songs usually want more delay than reverb.
Why does a vocal reverb sound fine solo and washy in the mix? The tail sits in the same 1 to 4 kHz region as consonants and guitars, so in the full mix it fills the gaps between words that intelligibility depends on. Filter the return, duck it against the dry vocal, and set the level in the mix, never in solo.
If your vocal sits wrong
Send me the chorus, dry vocal and instrumental as separate files, with the tempo. I will send back the vocal sitting in the track with the space printed, and the settings listed so you can rebuild them in your own session. Credits are at /credits, background at /about.