The exact dBFS targets, gain-staging order, and safety-track trick that keep dialogue clean, loud enough, and never clipped.
Audio levels are where amateur video gives itself away. Set them too low and you bury dialogue in hiss when you boost it later; set them too high and a single laugh clips into harsh distortion you can never undo. In the digital world levels are measured in dBFS (decibels relative to full scale), where 0 dBFS is the absolute ceiling and everything you record is a negative number below it. The targets are specific: aim for dialogue peaks landing around -12 to -6 dBFS and an average level near -18 to -12 dBFS, leaving comfortable headroom so unexpected loud moments don't hit zero. This guide walks the full gain-staging chain from mic to recorder, explains how much headroom to keep, when to use a limiter, and how a dual-channel safety track protects you from disaster on a one-take shoot.
dBFS stands for decibels relative to full scale. Unlike analog VU meters, the digital scale is anchored at the top: 0 dBFS is the maximum a digital system can represent, and any signal that tries to exceed it clips, producing permanent, unrecoverable distortion. Every level you record is therefore a negative number, and the bigger the negative number, the quieter (and farther from the clipping wall) the signal sits.
Because clipping is fatal but quiet recordings can be cleanly amplified later, you always record with the peaks well below 0. The space between your loudest peak and 0 dBFS is your headroom, the safety cushion for the unexpected loud moment. The whole game is recording hot enough to stay above the noise floor but cool enough to keep that cushion intact.
For spoken dialogue, set your gain so normal speech peaks around -12 to -6 dBFS, with the average (RMS) level sitting near -18 to -12 dBFS. That places loud peaks comfortably under the ceiling while keeping the voice well clear of the noise floor. A common professional convention is to aim for an average around -18 dBFS and let peaks reach roughly -10 to -6 dBFS.
The reason for the gap between average and peak is human speech dynamics: conversational delivery has wide swings between soft words and emphatic ones, plus the occasional laugh or raised voice. If you push the average too hot, those peaks have nowhere to go and clip. Music and music-bed material can sit a little hotter and more consistent, but for video destined for an edit, leave room; you will set final loudness (for example to platform targets near -14 LUFS for YouTube) in post, not on the recorder.
Gain staging means setting the signal level correctly at every point in the chain so nothing clips and nothing gets buried in noise. The most important stage is the first one: the preamp gain on your recorder, interface, or camera. Set this with the talent speaking at their real performance volume, not a quiet mic check, and dial gain up until peaks land in the -12 to -6 dBFS target, then stop. Do not crank gain to make the meter look full.
Avoid two classic mistakes. First, don't record too quietly and "fix it later": amplifying a weak signal in post raises the noise floor (hiss, hum, room tone) right along with the voice. Second, don't compensate for a too-quiet mic by maxing the preamp, which adds preamp self-noise; instead get the mic closer to the source. Each gain stage feeds the next, so a clean, well-placed level at the input gives every later stage room to work.
Headroom is the gap between your peaks and 0 dBFS, and it exists for the moments you can't predict: a sudden laugh, a cough, an emphatic line. A 6 dB safety margin is a reasonable minimum; many mixers keep more. If your peaks are hugging -6 dBFS and the talent gets animated, you can still clip, so when in doubt, back off a couple more dB. You can always raise a clean quiet recording; you cannot un-clip.
A limiter is a safety net that catches transients before they hit 0. Set a brick-wall limiter on the input (or use your recorder's built-in limiter) with a ceiling a few dB below zero, for example around -3 to -2 dBFS, so any peak that would otherwise clip is instead transparently held down. Use the limiter as insurance against the unexpected, not as a license to record dangerously hot; your underlying gain staging should still keep normal peaks in the target range on their own.
On a single-take shoot, an interview, or any moment you can't reshoot, the dual-channel safety track is the trick that saves the day. Record the same mic to two channels: the primary channel set to your normal target (peaks -12 to -6 dBFS), and a second "safety" channel recorded 6 to 12 dB lower. If the talent suddenly shouts or laughs and the primary channel clips, the quieter safety channel captures that peak cleanly, and you simply use the safety track for that moment in the edit.
Many field recorders and dual-system setups support this directly (sometimes called dual record or backup gain). If your recorder lacks it, use two separate inputs from a splitter, or record a second device. The cost is one extra channel of storage; the payoff is never losing an irreplaceable take to a single unexpected clip. Pair this with a limiter and conservative headroom and your audio is effectively bulletproof.
| Setting | Recommended | Why |
|---|---|---|
| Dialogue peak target | -12 to -6 dBFS | Loud enough above noise, with room under the 0 dBFS ceiling. |
| Dialogue average (RMS) target | -18 to -12 dBFS | Leaves the gap peaks need given natural speech dynamics. |
| Absolute ceiling | 0 dBFS (never reach it) | Any peak at 0 clips and is permanently distorted. |
| Minimum headroom | 6 dB below peaks (more is safer) | Cushion for unexpected laughs, coughs, and emphatic lines. |
| Preamp / input gain | Set with talent at real performance volume | First stage matters most; a quiet mic check under-sets gain. |
| Input limiter ceiling | -3 to -2 dBFS | Catches transients before they clip; insurance, not a crutch. |
| Safety track offset | 6-12 dB below primary channel | Captures clipped peaks cleanly on one-take, no-reshoot shots. |
| Mic distance | Get the mic closer to the source | Better fix than maxing preamp gain, which adds self-noise. |
| Avoid 'fix in post' low levels | Record hot enough to clear the noise floor | Boosting a weak signal later raises hiss and hum with it. |
| Final delivery loudness | Set in post (e.g. ~-14 LUFS for YouTube) | Loudness normalization is a mix decision, not a recording level. |
Set gain so normal speech peaks around -12 to -6 dBFS, with the average level near -18 to -12 dBFS. That keeps the voice well above the noise floor while leaving headroom under 0 dBFS for unexpected loud moments. You'll set the final delivery loudness (such as near -14 LUFS for YouTube) later in the edit, not on the recorder.
0 dBFS is the absolute digital ceiling. Any signal that reaches it clips, producing harsh distortion that can't be repaired. You always record peaks below zero with headroom to spare, because a clean, slightly quiet recording can be amplified later, but a clipped one is permanently ruined.
At least 6 dB between your peaks and 0 dBFS, and more if you can. Speech is dynamic, so a sudden laugh or raised voice can spike well above your normal peaks. If your peaks are already near -6 dBFS and the talent gets animated, back off a couple more dB to stay safe.
Yes, as a safety net. Set a brick-wall limiter with a ceiling around -3 to -2 dBFS so any transient that would clip is transparently held down instead. Treat it as insurance against the unexpected, not permission to record dangerously hot; your gain staging should still keep normal peaks in the target range.
It's recording the same mic to two channels: the primary at your normal target and a second channel 6-12 dB quieter. If the talent suddenly shouts and the primary clips, the quieter safety channel captures that moment cleanly, so you swap to it in the edit. It's essential insurance on one-take shoots you can't reshoot.
Audiences forgive a soft image far faster than they forgive bad sound. Here is how to choose the right mic for the way you actually shoot.
Camera, lens, lighting, audio and background for a clean talking-head YouTube studio that looks professional and stays simple to run solo.
A full-episode workflow that keeps the conversation natural, your brand consistent, and your clip pipeline full.
YouTube re-encodes everything you upload, so the goal is feeding it a clean, high-bitrate master it can compress without falling apart.
Once the footage is in the can, Studio432 turns it into something worth posting — concert films, music videos, gaming, vlogs and short-form, edited remote and worldwide. Send the footage, get a quote by email.
Get it edited →