VEIRA Clean Up

What does VEIRA Clean Up do?

It takes the background out of a recording: the room tone, the hiss of a preamp or a tape, the hum of a mains supply or an amp, a computer fan, air-conditioning, the traffic outside. One knob, REMOVAL, sets how much of it goes: 10 dB at 25 %, 20 dB at 50 %, 40 dB at 100 %. In the pauses the background goes down by exactly that; under the voice only what is clearly background goes, so the voice keeps its level and its tone. LISTEN plays only what is being taken out, and the BACKGROUND display shows the input, the cleaned output and the background the plugin has learnt.

How is it different from VEIRA Vocal Remover and VEIRA Gate?

VEIRA Vocal Remover takes the voice out of a finished song, for a karaoke track or an a cappella. VEIRA Clean Up does the opposite job: it takes the background out from under a voice. VEIRA Gate works on level alone: it mutes the track between the words and leaves it alone while the voice plays, so the background is still there under every word, and the quiet ends of words can get cut. VEIRA Clean Up works frequency by frequency, cleans under the words too, and never cuts a word short: breaths and the tails of words rise above its picture of the background and stay.

Do I need to take a noise print first?

No. The plugin learns the background by itself, in the pauses: the moments when the microphone hears only the room. A moment counts as a pause when it is within 6 dB of the quietest moment of the last three to six seconds and either 12 dB under the loudest or under −40 dBFS in itself. A pause of about a second is enough to start, and LEARNING lights on the display while it listens. The picture is refreshed at every pause and kept between them, so a fan that speeds up or traffic that builds is followed. A quiet background is learnt from the first moment; a loud one, over −40 dBFS, at the first pause after the voice begins, so with a street interview play from a little before the first line.

Nothing happens: REMOVING reads 0.0 and BACKGROUND reads --.

The plugin hasn’t heard a pause yet, so it has no background to remove. Play from a point where the recording has a second of background alone, the lead-in before the first word or a gap between phrases, and watch LEARNING light. If it never lights, the sound has no pauses: a held note, a strummed guitar or a dense mix never drops 12 dB under its loudest moment, and the plugin leaves it alone on purpose. Very quiet material, under −40 dBFS and hardly changing in level, is the other case: it is taken for background, so raise the level going in.

Where should I set REMOVAL?

Most voices sit well between 30 and 60 %: enough to take a home studio’s fan, computer and room out of the way (12 to 24 dB) without touching the voice. Turn it up until the background is gone from the pauses, then back a little if the voice starts to sound thin or watery. The quieter the background, the more you can take: a good recording with a little air can go to 100 %, while a loud background close under the voice (a camera microphone outside) sounds better at 40 to 55 % with a little of it left than hollowed out. Sung vocals take less than speech (25 to 55 %), because their air and reverb tails are part of the sound, and music takes the least (15 to 35 %). The presets are these starting points, one per situation.

What is LISTEN for, and can it end up in my bounce?

LISTEN plays the input minus the cleaned output: only what is being taken out. What you want to hear is the background alone, a steady hiss, a hum, the room. If you can make out words, consonants or the ring of a note, REMOVAL is too high for this recording; turn it down until they disappear, then switch LISTEN off and judge the voice itself. It is never saved in a preset, it can’t be automated, and it switches itself off when the host renders offline, so an offline bounce is always the cleaned sound. In a real-time bounce it stays as you left it, so switch it off first.

What is the latency, and can I record through it?

One analysis frame: 1,024 samples at 44.1 and 48 kHz (21 ms at 48 kHz), 2,048 at 88.2 and 96 kHz, 4,096 at 176.4 and 192 kHz. It is constant and reported to your DAW, which compensates for it on playback and in bounces, so the track stays in time. Bypassed, the plugin delays the sound by the same amount and keeps listening, so it still knows the background when you switch it back in. Compensation can’t make a live sound arrive early, and 21 ms is enough for a singer to notice: monitor the microphone directly while recording and use the plugin on playback. At REMOVAL 0 % the output is the input, bit for bit, only delayed.