Quiet and distorted are different problems
A recording that is too quiet can be turned up. A recording that clipped cannot simply be turned down to sound clean, because the tops of the waveform were never captured. Audacity defines clipping as a source sent to it louder than it can record, which chops off the tops and bottoms of the waveform. [1]
This guide is an editorial workflow built on Audacity's manual. It is not a microphone comparison or a restoration test, and it does not claim a universal input setting.
Which Audacity tool does what
| Tool | What it changes | Default | What it cannot do |
|---|---|---|---|
| Show Clipping in Waveform (View menu) | Draws vertical red bars where Audacity detects clipping | Off until enabled | It detects; it does not repair [1] |
| Normalize | Sets peak amplitude; can remove DC offset and balance stereo channels | -1 dB peak; Remove DC offset on; stereo channels independent off | Does not allow values above 0 dB and does not restore clipped peaks [2] |
| Clip Fix | Interpolates estimated peaks in clipped regions | Reduces amplitude by -9 dB to leave room for restored peaks | Only likely to be effective on lightly clipped audio; slow, so work on a few seconds at a time [3] |
| Loudness Normalization | Sets perceived loudness (LUFS) or RMS level, based on EBU R 128 | -23 LUFS perceived loudness, or -20 dB RMS | Matches loudness targets; it is not a distortion repair tool [4] |
Make a 30-second test take first
Audacity's test-recording tutorial suggests aiming for a maximum peak of about -6 dB, or 0.5 if the meter is set to linear, with the meter approaching but never touching the right-hand edge during the loudest parts. [1] Record normal speech including a naturally loud phrase, not only quiet narration, at your real microphone distance.
The tutorial's sequence is: enable silent monitoring from the Recording Meter, record a few seconds, adjust the slider in the Recording Meter, check the waveform for clipping, lower the level and try again if needed, then use Edit > Undo Record to discard the test. If the meter does not respond to the slider, which can happen with USB microphones, adjust the input volume in your system sound settings instead. [1]
Worked example: a fictional voice-over
In this invented case, a narrator's quiet opening is followed by one loud sentence. With clipping display on, that sentence shows red bars and sounds harsh. The editor applies Normalize at the default -1 dB: the whole clip changes level, but the harsh sentence still sounds damaged, which is what the manual implies, because Normalize changes peak level rather than reconstructing lost signal. [2]
The editor selects three seconds around the worst peak and tries Clip Fix on a copy. The manual only promises likely benefit on light clipping, so the editor compares the result by ear. [3] Because the narrator is available, the final decision is to record that one sentence again with the input lowered so peaks sit near -6 dB. The original take is kept and the replacement is documented. All of this is illustrative, not a measured result.
A decision checklist
- Red bars and audible harshness on the unprocessed source: re-record the passage if at all possible.
- Red bars but no audible damage on a brief peak: try Clip Fix on a copy of a few seconds and compare by ear.
- No red bars and audio is simply quiet: Normalize or Loudness Normalization is appropriate.
- Distortion appears only after your effects: inspect processing and combined track levels, not the source.
- Distortion on only one playback device: note it and test elsewhere before discarding the take.
- Change one variable at a time and keep an untouched original.
Choose the final level from the destination, not the tool default
Loudness Normalization defaults to -23 LUFS, which the manual identifies as the EBU standard. [4] That is a broadcast-oriented default, not necessarily what your podcast host or client specifies. Take the target from the actual brief or platform. Then play the export outside Audacity, including the loudest phrase and the transitions between speakers. If re-recording was impossible, label remaining distortion honestly rather than calling the audio restored.
Questions before you decide
Can Normalize fix clipping?
No. Audacity's Normalize sets peak amplitude and will not go above 0 dB. The manual does not describe it as repairing already clipped audio. For light clipping, Audacity offers Clip Fix; for heavy clipping, record again.
What recording level does Audacity recommend?
Its test-recording tutorial suggests a maximum peak of about -6 dB, or 0.5 on a linear meter, with the meter approaching but not touching the right-hand edge during the loudest parts.
What is the difference between Normalize and Loudness Normalization?
Normalize sets peak level, defaulting to -1 dB. Loudness Normalization targets perceived loudness in LUFS, defaulting to -23 LUFS based on EBU R 128, or an RMS level of -20 dB.
Sources & verification
Product details and prices can change. Check the linked provider before buying.
- Audacity manual: Tutorial - Making a Test Recording Accessed 29 Oct 2026
- Audacity manual: Normalize Accessed 29 Oct 2026
- Audacity manual: Clip Fix Accessed 29 Oct 2026
- Audacity manual: Loudness Normalization Accessed 29 Oct 2026
Sources link directly to providers. Product buttons may use separately labeled affiliate links. Read our disclosure.