How do I set recording levels so my audio does not clip?

Analogue tape distorted gracefully when pushed. Digital does not: above the maximum value there is simply nothing, so the waveform is flattened and the sound turns to harsh crackle that cannot be repaired.

Step by step

1. Understand the scale.Digital meters are marked in decibels below full scale, so zero is the absolute ceiling and everything usable is a negative number. Minus twelve is a healthy average level; minus six is loud; zero is failure.
2. Set the gain at the input.Gain is the control on the microphone preamp, interface or recorder, and it is the only one that matters for clipping. Turning down a fader afterwards does not undo a clipped input — the damage happened at the converter.
3. Test at performance level.Speak, sing or play at the loudest you will actually be, including the laugh, the emphasis and the sudden louder sentence. Setting levels on a quiet test and then performing normally is the commonest cause of a ruined take.
4. Leave headroom deliberately.Aim peaks at minus twelve and you have twelve decibels of margin for the unexpected. That margin costs you nothing, because raising the level afterwards in editing is free and lossless in a modern 24-bit recording.
5. Record at 24-bit if you have the option.It gives an enormous margin below the ceiling with no audible noise penalty, which is precisely what lets you record conservatively. At 16-bit the noise floor is higher and a very quiet recording is noisier when raised.
6. Turn off automatic gain control.Automatic level control raises the gain during quiet passages, which brings up the room noise and produces an audible breathing effect between phrases. Set the level manually and leave it.
7. Watch the meter, not the waveform.Peak meters show the highest instantaneous value, which is what clips. Some software shows a smoothed average that hides brief peaks entirely. If the recorder has a peak hold or clip indicator, use it.
8. Use a limiter as a safety net where available.Many recorders and interfaces offer a limiter or a dual-record safety track at a lower level. Both are insurance against a single loud moment destroying an entire take that cannot be repeated.
9. Normalise afterwards, not before.Once recorded cleanly, raise the whole file to the level you need in editing. For spoken content, aim at the loudness target of wherever it is going, which is usually around minus sixteen LUFS for podcasts and around minus fourteen for streaming platforms.

Tips

• If the meter is barely moving, do not fix it by raising the gain to maximum — move the microphone closer. Gain amplifies the noise floor as well as the voice.
• Record a thirty-second level test and listen to it on headphones before the real take. Levels judged by eye alone miss distortion that is already audible.
• Anything you cannot repeat — an interview, a performance, an event — deserves a second recorder set several decibels lower as insurance.

Common mistakes

• Recording as loud as possible to get a strong signal — It leaves no margin for the one loud moment, and clipping is permanent. Modern recording has plenty of dynamic range; the loudness is added afterwards.
• Turning the gain up instead of moving closer — Gain raises the preamp noise and the room sound equally with the voice. Halving the distance to the microphone gains signal without gaining noise.
• Leaving automatic gain control on — It constantly rides the level, so the background noise rises and falls between phrases. It is very audible and it cannot be undone.

If it doesn't work

Questions people ask