How to Set Recording Levels: The -18 dBFS Target and How to Hit It
Set recording levels so peaks land between -12 and -6 dBFS and the average sits near -18 dBFS RMS, which is the digital equivalent of 0 VU on analogue gear. Set the gain from the loudest phrase the performer will actually sing or play, not from a spoken level check, and engage the pad when the gain knob is already at minimum and the signal is still too hot. Twenty four bit recording gives about 144 dB of range, so headroom costs you nothing.
Definition: Gain staging is the practice of setting the level at each amplification point in a signal chain so that every stage runs in its linear range, with the first stage after the microphone doing almost all of the work.
There is exactly one knob in a home recording chain that sets a level you cannot undo, and it is the preamp gain. Everything downstream is either a fixed conversion or a fader that can be moved later. So the whole discipline reduces to a single question: where should the loudest moment of the performance land on the meter? The answer is peaks between -12 and -6 dBFS, average around -18 dBFS RMS, and the rest of this guide is why, and how to hit it without burning a take.
What do the numbers actually mean?
dBFS means decibels relative to full scale, where 0 dBFS is the largest number the converter can represent. Every value is therefore negative, and the scale has a hard ceiling with nothing above it. That is the crucial difference from analogue metering, where the red zone is a suggestion and going into it produces gradual saturation. In digital, hitting full scale produces clipped samples and there is no gentle version of it.
Peak is the highest instantaneous sample value in a passage. RMS is a time-averaged value that correlates far better with how loud something sounds. The difference between the two is crest factor, and it varies enormously by source: a bass guitar might run 8 to 12 dB of crest factor, a vocal 12 to 18, a close-miked snare drum 20 or more. That variation is why a single "record at this number" rule has to be stated as two numbers, one for peaks and one for average.
| Source | Typical crest factor | Peak target | Note |
|---|---|---|---|
| Spoken voice, close | 10 to 14 dB | -8 to -4 dBFS | Consistent, easy to set |
| Sung vocal, dynamic performance | 14 to 20 dB | -4 to -2 dBFS | Aim average lower, near -20 dBFS RMS |
| DI bass guitar | 8 to 12 dB | -10 to -6 dBFS | Watch open string attacks |
| DI electric guitar | 10 to 15 dB | -8 to -3 dBFS | Pick attack sets the peak |
| Acoustic guitar, condenser | 14 to 18 dB | -4 to 0 dBFS | Aim average near -20 dBFS RMS |
| Guitar amp, SM57 close | 10 to 14 dB | -8 to -4 dBFS | Very consistent once set |
| Snare drum, close | 18 to 24 dB | -6 to -3 dBFS | Pad usually required |
| Room mic on a kit | 16 to 22 dB | -8 to -4 dBFS | Set on the loudest fill, not the groove |
Read that table as one instruction: aim the average at -18 dBFS RMS, then check the peaks and make sure you have not run out of ceiling. If a source has huge crest factor, the average target moves down rather than the peaks moving up.
The interface preamp is where this happens, and its published gain range decides which microphones are practical. The Focusrite Scarlett 2i2 (4th Gen) has an Auto Gain button that listens to a few seconds of input and sets the knob for you, which is a genuinely good feature for anyone still learning to read a meter. The MOTU M4 puts a real LCD meter on the front panel, which matters more than it sounds: it is the only way at this price to set a level without looking at a screen. The gain staging calculator gives you the target for a specific source and crest factor.
Why -18 dBFS and not something louder?
Three separate reasons, and they stack.
Alignment. On an analogue console, 0 VU on the meter corresponds to +4 dBu at the output, and that point sits roughly 18 to 24 dB below where the electronics start to distort. When digital recording arrived, the broadcast world had to decide what 0 VU should equal in dBFS, and the European standard settled on -18 dBFS while the American film standard settled on -20 dBFS. That choice is still baked into everything. Every analogue-modelled plugin, every console emulation, every tape saturation plugin is calibrated so that its sweet spot lands with roughly -18 dBFS at the input. Feed one a signal peaking near zero and you are driving a virtual 1176 twenty decibels harder than its designer intended.
Headroom. The loudest thing a performer does is almost never the thing they did during the level check. Leaving 12 to 18 dB of space above your average means the chorus that arrives 30 dB louder than the verse does not clip. In digital, one clipped sample in the best take of the day is not repairable in any meaningful sense.
Bit depth makes it free. Twenty four bit gives about 144 dB of theoretical dynamic range against 96 dB for sixteen bit. If you peak at -18 dBFS you are notionally giving up 3.0 bits and still working with about 126 dB. No microphone, no preamp and certainly no home room comes anywhere near that. Peaking at -6 dBFS instead only recovers 1.0 bit of theoretical range you were never going to use. The dB reference levels chart lays the whole scale out side by side.
The lesson that made this permanent for me was a session where I set gain off a singer saying "check, check" into the microphone, got a nice healthy meter, and rolled. She opened her mouth and the first chorus clipped in about four places. Not near the ceiling, actually clipped, because a belted phrase runs fifteen to twenty decibels above a spoken one. That take had a quality in it I never got back over the next hour of trying. Now I ask for the single loudest moment in the song before I touch the knob, and if the singer is not warmed up enough to give it to me, I set conservatively and record at a level that looks too quiet. Nobody has ever complained about a take being four decibels low.
How do you set the gain, step by step?
- Get the microphone in position first. Distance changes level by 6 dB every time you double it, so setting gain before the microphone is where it will be is setting it for a different session.
- Turn the gain to minimum, put phantom power on if it is a condenser, and get the headphone level up so the performer can hear themselves.
- Ask for the loudest section of the actual song. Not a count-in, not a spoken check, not the first verse. The biggest moment in the arrangement, at the intensity they will actually deliver it.
- Bring the gain up until peaks read around -8 dBFS. Use the interface meter or the DAW input meter; they should agree within a decibel.
- Check the RMS. If your DAW shows an averaging meter, the sustained level should sit around -18 to -20 dBFS. If the peaks are right and the average is 30 dB below them, you have a very spiky source and the peaks are worth backing off further.
- Do not touch it again mid-take. A gain change halfway through a take produces an edit point you cannot hide.
One thing that is not on this list: recording with a compressor already engaged. Some interfaces, such as the Universal Audio Volt 276 , put a hardware compressor in front of the converter, and it can be the right call for a wildly dynamic singer. But it is a decision you cannot undo, so use it to catch peaks rather than to shape the sound, and know that committing is a workflow choice rather than a technical improvement.
When do you need the pad, and when do you need more gain?
These are the two ends of the same problem, and the diagnosis is simple. If the gain control is at or near minimum and the signal is still too loud, the input stage is being overdriven before the gain control acts, and only a pad fixes it. A pad is a fixed attenuator, usually 10 or 20 dB, placed ahead of the preamp. The Audio-Technica AT2035 has a 10 dB pad on the microphone body, which is the better place for it because it protects the microphone's own electronics as well as the preamp's.
The opposite case is a low output dynamic. The Shure SM7B publishes a sensitivity of -59 dBV per pascal, which is about 1.12 millivolts at a normal speaking level, and it wants 60 to 65 dB of clean gain. Fourth generation Scarlett preamps publish 69 dB, which clears it. Plenty of budget interfaces stop near 50, and running one flat out is exactly where its own hiss becomes part of your recording. An inline booster fixes it: a Cloudlifter CL-1 mic activator, with cable adds about 25 dB and a Triton Audio FetHead in-line preamp about 27, both drawing phantom power to do it. Condenser vs dynamic microphone works through the sensitivity gap in detail.
A third case that confuses people: a hot instrument input. A DI'd active bass or a keyboard output can be far hotter than a microphone, and plugging a line level source into a microphone input without engaging the instrument or line setting is a common way to get distortion with the gain at minimum.
What about 32-bit float, and does it end the argument?
Thirty two bit float recording is a real change rather than a marketing number. A float format stores an exponent alongside the value, so the usable range is enormous and a signal that exceeds what would be full scale in a fixed point format can be brought back down afterward without clipped samples. Interfaces such as the Zoom UAC-232 (32-bit float) implement it with two converters running at different levels, and the recorder picks whichever one is not overloaded.
What it fixes: the digital ceiling. What it does not fix: the analogue stage. If the preamp or the microphone capsule itself is being overdriven, the distortion has already happened before the converter sees it, and no amount of floating point saves you. So 32-bit float removes one of the three reasons to leave headroom, and leaves the other two standing. Set gain the same way.
The habit I would give anyone starting out is to leave the DAW meters in peak-hold mode and look at them once, at the end of every take, rather than watching them during it. Watching the meter while someone is performing is the fastest way to stop listening to the performance, and the performance is the only part of this you cannot fix later. I do a lot of work to deadline for picture, and the takes that get used are almost never the technically cleanest ones. They are the ones where the person in the room was not being managed. Set the gain conservatively, then stop looking at the numbers.
Where else in the chain does level matter?
Inside the DAW, much less than people think. Once the signal is in a 32-bit float mixer, an individual channel fader is not degrading anything, and there is no meaningful noise penalty to pulling a track down 12 dB. The two places it does matter are analogue-modelled plugins, which expect roughly -18 dBFS at their input to behave as designed, and the master bus, which should have headroom left for whatever happens at the mastering stage.
On the way back out, the monitor level control on the interface is not part of the recording chain at all: turning your monitors down does not change the file. That sounds obvious and yet it is the single most common confusion for people learning the chain. The full path is laid out in home studio signal flow explained, and the vocal-specific version of everything on this page is in how to record vocals at home.
Related tools
- Gain staging calculator: peak and RMS targets for a specific source
- dB reference levels chart: dBFS, dBu, dBV and VU side by side
- Best audio interface for a home studio: the preamps that have the gain range to do this
Frequently asked questions
What level should I record at in dBFS?
Peaks between -12 and -6 dBFS with the average sitting around -18 dBFS RMS. That gives you roughly 12 dB of crest factor for a typical vocal and leaves at least 6 dB of headroom above the loudest thing the performer does. It is not a rule about the file sounding better; it is a rule about the analogue stage in front of the converter running in its comfortable range and about surviving a louder second take.
Why is -18 dBFS the standard reference level?
Because it aligns digital to the analogue gear that came before it. On an analogue console 0 VU corresponds to +4 dBu, and that meter sits well below the point where the electronics distort. When digital arrived, the broadcast standard set 0 VU equal to -18 dBFS so an engineer moving between the two saw the same needle position mean the same thing. Every analogue-modelled plugin still expects roughly that level at its input.
Does recording quietly hurt the sound in 24-bit?
Practically no. Twenty four bit gives about 144 dB of theoretical dynamic range. If your peaks land at -18 dBFS you are still working with about 126 dB, which is far more than the noise floor of any microphone, preamp or room will let you use. The limiting noise source in a home studio is the room and the preamp, never the bit depth. Headroom is free, so there is no reason to track hot.
When should I use the pad on a microphone or interface?
When you have the gain knob at or near its minimum and the signal is still too loud. That means the preamp input stage is being overdriven before the gain control ever gets to act. A 10 or 20 dB pad in front of it fixes what turning the knob down cannot. Typical cases are a kick drum, a snare at close range, a cranked guitar cabinet, or a loud singer on a high output condenser.
How do I set gain for a singer without wasting a take?
Have them perform the loudest section of the actual song, at performance intensity, not speak a level check. A spoken check runs 10 to 20 dB below a belted chorus, so gain set from speech clips on the first real phrase. Ask for the biggest moment in the song, set peaks to about -8 dBFS, then roll the take. The habit costs thirty seconds and saves the first take, which is often the best one.
What is the difference between clipping the preamp and clipping the converter?
The converter clips when the sample value reaches full scale, and that is an absolute wall producing hard digital distortion. A preamp clips when its own analogue output stage runs out of voltage swing, which can happen before the converter meter shows anything if the gain structure is wrong. If your meter reads -10 dBFS but the sound is already crunchy, suspect the preamp or a missing pad rather than the converter.
Working out your own room and signal chain? The Home Studio Build Planner is the paid version of these pages: 8 printable worksheets you fill in with your own numbers, plus the full PDF, $29.