Levels: gain, clipping and headroom
Clipping is the one mistake you cannot fully undo. Audacity's Clip Fix tries to rebuild clipped peaks but is only likely to help with lightly clipped audio, and it lowers the level by 9 dB by default to make room. [7] Record conservatively instead.
| Term | Plain meaning | What to do |
|---|---|---|
| Gain | How much a signal is amplified, at the input or in the editor | Set input gain so speech peaks around -6 dB on the meter [1] |
| Clipping | The signal is louder than the system can record, so waveform tops and bottoms are cut off | Turn on View, Show Clipping in Waveform in Audacity; if red bars appear, lower the level and re-record [1] |
| Headroom | The space between your loudest peak and 0 dB | Leave it while recording; you can always boost later [1] |
| Peak normalization | Raises or lowers a track so its loudest sample hits a target | Audacity Normalize defaults to -1 dB and cannot go above 0 dB [2] |
| Amplify | Applies the same gain to everything selected, keeping balance between tracks | Use when tracks are already balanced; Audacity Amplify can push into clipping [2] |
| DC offset | The waveform is not centered on zero, which can cause clicks and waste headroom | Leave Remove DC offset ticked in Audacity Normalize [2] |
Loudness: LUFS, LKFS and true peak
LUFS, loudness units relative to full scale, measures how loud audio sounds rather than how high its peaks are. Audacity's Loudness Normalization is based on EBU R 128, and Apple states its podcast target in LKFS, measured with true peak per ITU-R BS.1770-5. [3][4]
Apple pairs its loudness target with a true-peak ceiling and warns that audio might clip if the recommended true-peak value is not respected. [4]
- Action: for a podcast-style video on Apple Podcasts, normalize to -16 LKFS and check true peak is at or below -1 dBFS. [4]
- Action: do loudness normalization as the final step before export; Audacity's manual says adjusting maximum amplitude is best done last. [3]
- Action: Audacity treats mono as dual-mono by default, unlike some LUFS meters that read mono 3 LU lower; turn that option off if you need to match those meters. [3]
| Context | Loudness target | Peak limit | Source |
|---|---|---|---|
| Apple Podcasts | About -16 LKFS, plus or minus 1 dB | True peak no higher than -1 dBFS | [4] |
| EBU broadcast | -23 LUFS | Not stated on the Audacity page | [3] |
| US television | About -24 LUFS | Not stated on the Audacity page | [3] |
| Audacity Loudness Normalization default | -23 LUFS | Not applicable | [3] |
| Descript auto-leveling | Around -16 dB per clip, units as stated by Descript | Not stated | [9] |
Noise: noise floor, noise reduction and gating
Gate parameters in plain words: threshold is the level that opens or closes the gate, level reduction is how much quiet audio is turned down, attack is how fast it opens, hold is how long it stays open after the signal drops, and decay is how gradually it closes. [5] Descript turns room tone off for clips whose file has Studio Sound enabled, so pick one approach per clip. [8]
| Term | Plain meaning | What to do |
|---|---|---|
| Noise floor | The level of background noise at each frequency when nobody is speaking | Record a few seconds of silence so tools can learn it [6][8] |
| Noise profile | A sample of noise only, used to calculate what to remove | Audacity needs at least 2048 samples, about 0.05 s at 44.1 kHz; longer is better [6] |
| Noise reduction artifacts | Short random tones, sometimes called musical noise, left by over-processing | Take a profile that represents the noise across the whole track; Audacity says artifacts can appear if Sensitivity is too low. Defaults are 6 dB reduction and 6.00 sensitivity [6] |
| Noise gate | Turns down audio that falls below a threshold, such as hiss between sentences | Start at a modest level reduction; deep settings cause pumping [5] |
| Room tone | The natural background sound of the room, used to fill silences so edits do not sound dead | Record at least 15 seconds of silence before speaking for Descript's room tone [8] |
Dynamics: compression and auto-leveling
| Term | Plain meaning | What to do |
|---|---|---|
| Compressor | Reduces dynamic range so quiet and loud parts are closer together | Use before final loudness so speech sits evenly [10] |
| Threshold | Level above which compression starts | Set just below normal speech peaks [10] |
| Ratio | How strongly the signal above threshold is reduced; at 10:1, 10 dB over becomes 1 dB over | Start gentle for voice [10] |
| Attack and release | How quickly compression starts after the threshold is crossed, and how quickly it lets go | Audacity notes output can briefly overshoot on attack and undershoot on release; listen on headphones [10] |
| Make-up gain | Gain added after compression to restore level | Then normalize loudness last [10][3] |
| Auto-leveling | Automatic per-clip loudness matching | Descript's version is on by default; split around spikes such as a sneeze and re-run [9] |
Format: sample rate and bit depth
Sample rate is how many times per second audio is measured; 48 kHz is the video standard YouTube recommends for uploads. [11] Bit depth is how finely each sample is measured. Apple accepts WAV or FLAC at a minimum of 44.1 kHz at 16 or 24 bit and recommends 24 bit; it requires stereo, or two identical channels if the source is mono. [4]
Action: record and edit at 48 kHz for video, so you never resample.
A processing order that avoids most problems
- Record with peaks around -6 dB and check for clipping. [1]
- Capture 15 to 30 seconds of room tone at the start. [8]
- Remove noise with a clean noise profile that represents the whole track, starting from Audacity's defaults. [6]
- Gate gently if hiss remains between phrases. [5]
- Compress for even speech, adding make-up gain. [10]
- Normalize loudness last to your platform target, and keep true peak at or below -1 dBFS for Apple Podcasts. [3][4]
Questions before you decide
What is the difference between normalize and loudness normalization?
Audacity's Normalize sets peak amplitude, by default to -1 dB, while Loudness Normalization targets perceived loudness in LUFS, by default -23 LUFS based on EBU R 128. Two clips with the same peak can still differ in loudness.
What loudness should a video podcast on Apple Podcasts use?
Apple asks for overall loudness around -16 LKFS with a tolerance of plus or minus 1 dB and a true peak no higher than -1 dBFS, measured per ITU-R BS.1770-5.
What recording level should I aim for?
Audacity's test recording tutorial suggests about -6 dB on the meter for the loudest parts, leaving room below the clipping point. You can boost the level later in editing.
Why does my noise reduction sound watery?
Audacity calls these artifacts musical noise and says they can happen when the noise floor is higher than was estimated, for example because Sensitivity is set too low or the noise profile does not represent the whole track. Use a representative noise-only sample and check the Sensitivity setting.
Sources & verification
Product details and prices can change. Check the linked provider before buying.
- Audacity Manual: Making a test recording Accessed 30 Oct 2026
- Audacity Manual: Normalize Accessed 30 Oct 2026
- Audacity Manual: Loudness Normalization Accessed 30 Oct 2026
- Apple Podcasts for Creators: Audio requirements Accessed 30 Oct 2026
- Audacity Manual: Noise Gate Accessed 30 Oct 2026
- Audacity Manual: Noise Reduction Accessed 30 Oct 2026
- Audacity Manual: Clip Fix Accessed 30 Oct 2026
- Descript Help: Room tone Accessed 30 Oct 2026
- Descript Help: Auto-leveling Accessed 30 Oct 2026
- Audacity Manual: Compressor Accessed 30 Oct 2026
- YouTube Help: Recommended upload encoding settings Accessed 30 Oct 2026
Sources link directly to providers. Product buttons may use separately labeled affiliate links. Read our disclosure.