Vocal Recorded Too Quiet? How to Fix It Without Raising the Noise
Recording · 7 min read
A vocal recorded too quiet is not a lost vocal. In a 24-bit session, raising a take that peaks at −30 dBFS by 20 dB adds no digital noise worth hearing; what it raises is the room and preamp hiss already in the file, by the same 20 dB. So the job is not "make it louder" but "make it louder while keeping the noise where it was": clip gain first, noise cleanup second, compression last, and an honest look at the gap between voice and hiss before any of it, because under 30 dB the fastest fix is a second take.
The order that keeps noise down
Clip gain first
Measure the gap · raise the take · clean the gaps · then compress
The 60-second answer
Read the meter on a silent gap: that is your noise floor. Read it on the loudest phrase: that is your peak. 40 dB or more apart and you are fine. Clip-gain the take so the loudest peak lands near −8 dBFS, then level the phrases to within ±3 dB with per-region clip gain. Run noise reduction next, learned from that gap, 6–8 dB and no more. High-pass at 80–100 Hz. Only then compress, 3:1 with 3–5 dB of gain reduction, and add an expander after it if hiss is still audible between words. If the gap was under 30 dB at the start, every step makes the hiss more obvious; re-recording with the gain 15 dB higher is faster than fighting it.
Why raising the fader is not the problem
People worry that turning a quiet recording up "adds noise". In a 24-bit file it does not, at least not noise you can hear. The quantization floor sits far below anything a home interface produces, and a bedroom with a computer fan is louder than the interface by a wide margin. So the noise that matters was captured by the microphone: the fan, the street, the preamp hiss at whatever gain you used. Raising the clip raises the voice and that noise together. The ratio between them was fixed the moment you pressed record, and nothing after that improves it except noise reduction, which trades hiss for artifacts.
What does make it worse is compression. A compressor pulling 8 dB of gain reduction with 8 dB of makeup leaves the loud words where they were and lifts the quiet gaps by 8 dB, hiss included. That is why order matters more than any single setting: everything that raises the gaps comes after everything that cleans them. The headroom article covers the levels to aim for next time.
Measure the gap before you touch anything
Loop a two-second silence between phrases and read the meter. Loop the loudest chorus line and read the peak. Subtract. Above 40 dB, treat the take as normal and skip noise reduction unless you hear hiss in the mix. Between 30 and 40 dB, follow every step below; it will sound good but you will work for it. Under 30 dB, the hiss sits roughly where a reverb tail would, under every held note, and noise reduction strong enough to remove it leaves the vocal underwater. Re-record, with the gain set so a loud phrase peaks around −12 to −6 dBFS; the Gain Staging Calculator gives you the number. For a quick listen before you start, the Audio Normalizer will peak-normalize the file, but it does not replace per-phrase clip gain and it raises the noise by the same amount.
Rescuing a quiet vocal step by step
| # | The move |
|---|---|
| 01 |
Read the two numbers Peak of the loudest phrase, level of the quietest gap. This decides whether steps 02–06 are worth doing. Gap ≥ 40 dB: easy · 30–40: workable · < 30: re-record |
| 02 |
Clip-gain the whole take to a sane peak One gain move on the region so the loudest word peaks near −8 dBFS. No plugins yet. Peak target −10 to −6 dBFS |
| 03 |
Level the phrases by hand Split at phrase boundaries, bring quiet lines up and loud lines down, so the compressor later has 5 dB of work, not 15. The Vocal Clip Gain Planner writes the move list. Phrases within ±3 dB |
| 04 |
Noise reduction, learned from a real gap Train the profile on at least one second of silence, apply 6–8 dB. If words start to sound hollow, you went too far. 6–8 dB reduction · learn from a 1 s gap |
| 05 |
High-pass, then compress gently HPF removes the rumble that would otherwise pump the compressor. Keep the ratio moderate so makeup gain stays small. HPF 80–100 Hz · 3:1 · 3–5 dB GR · makeup to match |
| 06 |
Expand the gaps after the compressor A downward expander, not a hard gate, pulls the hiss between words down without chopping breaths and word endings. Expander 1:2 · range −10 dB · release 150 ms |
The trick: two light noise passes instead of one heavy one
Every noise reduction algorithm starts producing a watery, chirping artifact past a certain amount of reduction, usually somewhere between 8 and 12 dB depending on the tool. A single 12 dB pass sits right on it. Two passes of 5–6 dB each, with the profile re-learned between them, remove about the same total hiss and stay under the artifact threshold both times, because the second pass works on a cleaner signal with less to misjudge. Put the high-pass filter between the passes so the second is not spending effort on rumble. Then, instead of more reduction, use automation: ride the clip gain down 4–6 dB in every gap longer than half a second. Manual rides are artifact-free by definition. The gate and noise cleanup guide goes deeper on the expander settings.
Chains with the reasons left in.
The Vocal Chain Bible is 88 real vocal chains from hit records, as the engineers described them — mic, preamp, every plugin and setting, in order, with a free-plugin version of every chain. Every chain comes with the reason behind each setting, so you know what to move for your voice.
Get the Vocal Chain Bible — $48 →Mistakes that kill it
- Compressing first to "bring it up". The compressor lifts every gap by the makeup gain. Clean the gaps before anything that raises them.
- Normalizing instead of clip-gaining. Peak normalization sets one level for the whole file. A quiet verse and a loud chorus need different amounts, and the compressor pays for the difference with noise.
- One 15 dB noise reduction pass. It removes the hiss and the consonants together. Stay at 6–8 dB per pass.
- Using a hard gate. It cuts the ends of words and makes the hiss come and go, which is worse than steady hiss. Use an expander with a limited range.
Keep going
FAQ
Does turning up a quiet vocal add noise?
In a 24-bit session, not digital noise you can hear. It raises the room and preamp noise already in the recording by the same amount as the voice. Whether that is acceptable depends on the gap you measured at the start.
How quiet is too quiet to fix?
If the loudest phrase is less than about 30 dB above the silent gaps, noise reduction strong enough to hide the hiss will damage the voice. Re-record with the gain 10–15 dB higher and the mic 5 cm closer.
Should I normalize a quiet vocal before mixing?
Only as a quick way to hear it at a usable level. Per-phrase clip gain does the real work, because it evens out loud and quiet sections so the compressor and its makeup gain stay small.
Which comes first, noise reduction or EQ?
Noise reduction on the raw take, then the high-pass filter, then corrective EQ, then compression. With two light noise passes, put the high-pass between them.
Who's behind this
Ivan Meshcheriakov · IvanFromRnD
5+ years ago I moved to LA with zero connections and started building from nothing — 14-hour days, meeting people and interning at studios. It took me 5+ years to meet and interview 100+ Grammy-winning engineers — then I turned it all into the Vocal Chain Bible.
Read my full story →