What is the right workflow for standardizing audio clips?

Occasionally, I like to record how I play certain pieces on the piano so that I can listen to them again. My main problem is that the levels at which the volume peaks can be inconsistent between pieces. To address this and improve overall quality, I would like to standardize any new recordings before importing them into my library.

Here’s the workflow I came up with so far:

  1. Remove ‘DC offset’
  2. Apply ‘high pass filter’ (20-30Hz, 24dB)
  3. Apply ‘noise reduction’ (9dB reduction, 6.00 sensitivity, 3 smoothing)
  4. Add fade-in / out effects at the ends
  5. Adjust the amount of silence at the ends (how much?)
  6. Compress outliers (podcast / radio preset?)
  7. Normalize (-2dB)

Can anyone offer any feedback or guidance? I’m new to audio engineering and I have little to no experience with Audacity and its functions / tools.

There is no “standard” (except for ACX audiobooks).

Most of the “quality” comes from the quality of the recording and the performance. That includes a good microphone and a good room (either a “dead” soundproof studio or a music hall with “good acoustics”). It’s not easy to get a good acoustic piano recording at home. :frowning: … Or any acoustic instrument, really.

There is The Loudness War where everybody tries to be louder than everybody else. :stuck_out_tongue: That makes everything constantly-loud, killing the dynamic contrast and making music boring!

Radio stations use compression and limiting to make everything “loud” and the same volume, and they try to be louder than the other stations.

All of the popular streaming services use loudness normalization (different from regular peak normalization) so there isn’t much variation between different recordings. They mostly do this by lowering the volume of the louder tracks. Some quiet sounding tracks still can’t hit the target volume without clipping and they will remain below the target. The streaming services don’t use compression/limiting so unlike broadcast radio they aren’t altering the “performance”.

Your normalization step will set the peaks to exactly -2dB. You can set it to 0dB to make it 2dB louder. …Nothing bad happens when you get close to 0dB but you can get clipping (distortion) if you go over. That’s the highest peak(s) in the file and it’s possible that there is only one quick-short peak at that level. Regular normalization is a simple-linear volume adjustment and it doesn’t affect sound character-quality unless you go into clipping.

Note that peaks don’t correlate well with loudness. Almost all commercial recordings are compressed or limited (limiting is a fast-kind of dynamic compression) to bring-up the overall loudness without boosting/clipping the peaks.

Compression changes the dynamics (the relationship between loud and quiet parts) so it’s creative-judgement call. If overdone it can mess-up the sound.

In general, a solo acoustic recording won’t be as loud as a more “dense” recording of a full band, etc.

If your radio preset isn’t getting it loud enough try the legacy limiter (with make-up gain). Effects → Legacy → Legacy Limiter. Maybe start with a limit of -6dB and you can experiment from there.

Also note that anything that boosts the volume will bring-up the background noise making it more noticeable.

…Most of us don’t have the skills (or tools) of a professional mastering engineer and we can’t get the same loudness with as little damage. But if you have a good stereo with enough analog power & gain, you can simply turn-up the playback volume. :wink:

Hi! Thank you for your response!

It may be a character flaw, but I firmly believe in order and process. Realistically, I will probably just use the normalization feature to make everything sound more consistent, however good or bad my equipment may be. Nevertheless, I would appreciate more detail on the other steps for when they are necessary. From what I understand, normalization only makes sense to use when the track has a consistent level of loudness and there aren’t any spikes - that’s where compression is supposed to come in.

Also, what is a ‘limiter’? Is it another step I should include?

[/quote]

Almost all commercial recordings are normalized and there is no harm. It’s almost always a good last-step.

…GoldWave (a non-free “competitor” to Audacity) calls it “maximize” which is a better English word but “normalize” is the correct audio terminology. Again, it’s a linear volume adjustment that doesn’t affect character or quality. It’s like adjusting the volume control before playback starts.

With an album some songs may be intentionally louder than others and in that case the album will be normalized as a whole so the relative loudness between songs isn’t changed.

…With the default settings, Audacity’s Amplify effect will also normalize for 0dB peaks.

I wouldn’t say “supposed to”. Since it changes the sound it’s optional. Real live music isn’t compressed. :wink:

In general, dynamic compression reduces the dynamics by making the loud parts quieter and/or the quiet parts louder.

In practice, compression usually lowers the loudest parts and then “make-up gain” (or normalization) is applied to make everything louder.

Real live music has intentionally louder & quieter parts… But it’s usually not practical (or desirable) to have the full dynamic range of an orchestra or rock band (or maybe even a piano) in your living room or car and most stereos don’t go loud enough without distorting the loud parts. (In a car, the quiet parts can get lost in road noise and you end-up constantly adjusting the volume control, etc.).

Limiting is a fast kind of compression that “pushes down” the peaks without hard-clipping. Again it’s normally used with make-up gain to bring-up the loudness without distortion.

Limiting often gives better results than regular compression and there are fewer settings to mess with (or mess-up).

BTW - Automatic volume control (or automatic gain control or leveling) are a slow-kind of compression. They usually don’t work very well with music but they can sometimes work with speech.

All very useful information, but I still need a more detailed breakdown. I just finished writing a guide on the topic - what do you think of it?

Step 1: Remove DC offset

Details

  • Where: Effect > Volume and Compression > Normalize
  • Settings: Remove DC offset only

Step 2: Apply high-pass filter

Details

  • Where: Effect > EQ and Filters > High Pass Filter

Description

  • Frequency: The threshold below which sound is attenuated (20 Hz)
  • Roll-off: Decrease in amplitude for each octave below threshold (12 dB)

Frequency

Description Setting Use cases
Inaudible 20 Hz Music tracks
Background 80 Hz to 100 Hz Speech / vocals
Foreground Up to 150 Hz Poor recording

Roll-off

Description Setting Use cases
Gentle 12 dB Light cleanup
Moderate 24 dB Muffling
Steep 48 dB Abrupt cut-off

Step 3: Apply noise reduction

Details

  • Where: Effect > Noise Removal and Repair > Noise Reduction
  • Step 1: Get Noise Profile
  • Step 2: Apply

Description

  • Reduction: Amount of volume reduction applied to identified noise (6 dB)
  • Sensitivity: How much of the audio is considered noise from 0 to 24 (6.00)
  • Smoothing: Number of neighboring frequency bands to adjust (6)

Reduction

Description Setting Use cases
Low 6 to 12 dB Mild noise
Moderate 12 to 20 dB Background noise
High 20 to 30 dB Foreground noise

Sensitivity

Description Setting Use cases
Low 6 to 12 Constant hum
Moderate 12 to 18 Diverse noises
High 18 to 24 Very distorted

Smoothing

Description Setting Use cases
Light 0 to 3 Music tracks
Balanced 3 to 6 Mixed content
Aggressive 6 to 10 Spoken word

Step 4: Adjust silence

Details

  • Retain clip: Edit > Remove Special > Trim Audio
  • Nullify clip: Edit > Remove Special > Silence Audio
  • Add silence: Generate > Silence

Step 5: Add fade-ins / outs

Details

  • Where: Effect > Fading > Fade In / Out

Step 6: Truncate silence

Details

  • Where: Effect > Special > Truncate Silence

Description

  • Threshold: Level below which audio is treated as silence, between -20 and -80 dB
  • Duration: A clip’s minimum duration below the threshold to be treated as silence

Threshold

Description Setting Use cases
Very low -70 dB and below Complete silence
Low -60 dB Most tracks
Moderate -50 dB Ambient noise
High -40 dB and above Background noise

Duration

Description Setting Use cases
Short 0.5 to 1s Start (standard)
Longer 1 to 2s End (standard)
Long 2s or more End (classical, jazz)

Step 7: Apply compression

Details

  • Where: Effect > Volume and Compression > Compressor

Description

  • Threshold: Level above which compression is applied
  • Ratio: Amount of compression applied to the targeted areas
  • Knee width: How gradual the compression curve around the threshold is
  • Lookahead: Compression is applied gradually before the input exceeds the threshold
  • Attack: Compression is applied gradually after the input has exceeded the threshold
  • Release: Compression is released gradually after the input has gone back below the threshold
  • Make-up gain: Amount by which the track is amplified after compression

Threshold

Description Setting Use cases
High -1 to -10 dB Light retouching
Moderate -10 to -20 dB Handling spikes
Low -20 to -30 dB Boosting bass

Ratio

Description Setting Use cases
Mild 2:1 to 3:1 Balancing voices
Moderate 4:1 to 5:1 Creating consistency
Heavy 6:1 or more Loud sound effects

Knee width

Description Setting Use cases
Hard 1 to 3 dB Percussion
Softer 3 to 5 dB Most music
Gentle 5 dB or more Soft instruments

Lookahead

Description Setting Use cases
Short 0 to 3ms Natural timing
Longer 3 to 10ms Abrupt spikes

Attack

Description Setting Use cases
Short 0 to 10ms Percussion
Medium 10 to 20ms Most music
Long 20 to 30ms Soft instruments

Release

Description Setting Use cases
Short 50 to 100ms Percussion
Medium 100 to 200ms Most music
Long 200 to 300ms Soft instruments

Step 8: Normalize track

Details

  • Where: Effect > Volume and Compression > Normalize

Description

  • Peak amplitude: The amplitude of the track is scaled based on this value

Peak amplitude

Description Setting Use cases
Standard -1.0 dB Most audio