Replay Gain - question on odd observation

I have used Replay Gain to great effect since being advised of it here. Many thanks. My project is to string together songs, all mixed into one another. When I import a song into my project, I double click on it and use Replay Gain. My understanding is that this sets the volume to the same level for all songs I use Replay Gain on. On a project recently, I found that one song was louder than the others despite using Replay Gain. I actually recall this for another project I once did. Can you shed some light on this?


If you use the ReplayGain Analyze option, do they all give the same result? (You can select/highlight one song at a time even after combining.)

It could be a couple of things… Some quiet-sounding songs can’t be adjusted-up to the ReplayGain target without clipping so if you don’t allow clipping they won’t be completely volume matched. And, you probably don’t want clipping (distortion).

Or it could just be a matter of perception. ReplayGain tries to take human perception into account but two different people might not agree when two different songs are volume-matched, especially if they are different styles/genres or if one song is highly dynamic, etc.

My project is to string together songs, all mixed into one another.

If you have a limited-manageable number of songs (like if you’re making a CD with 10 or 20 songs) it’s better to do it manually by-ear -

  1. Normalize all of the songs (for 0dB peaks, or near 0dB peaks).
  2. Listen to all of the songs and if they are not volume matched, choose the quietest one as your reference. (You’ve already normalized so you can’t go louder without clipping.)
  3. Adjust the louder songs down (by ear) to match the reference.

i ‘analyzed’ all the tracks…all but two showed ‘minus 3’ as a result. One showed -2.9 and another -3.1.

I’m noticing that very sparse piano mainly tracks sound quieter. Very sparse vocal only tracks (minimal instrumentation) is even louder. More busy arrangements are quieter.

i ‘analyzed’ all the tracks…all but two showed ‘minus 3’ as a result. One showed -2.9 and another -3.1.

Good news/bad news…

ReplayGain is working and (according to it’s algorithm) it has matched the volumes to +/- 0.1dB. If your ears/brain disagree you’ll have to adjust manually by-ear. :wink:

Can I use the ‘amplify’ effect to raise volume on a particular track? I could give each song a seperate Stereo track (and use its volume adjustment) in Audacity but am trying to avoid. I just flipped between two tracks, alternating songs between the two.

Can I use the ‘amplify’ effect to raise volume on a particular track? I could give each song a seperate Stereo track

Yes, and you don’t have to make “separate tracks”. As long as there is silence between tracks you can just select/highlight the audio from one track at a time. (You don’t want any sudden jumps & or down in volume so the end-points should be during silence.)

Just watch out for clipping. You may have to make the loud songs quieter rather than boosting the quiet-songs.

thanks…can you tell me what these specs mean when I use ‘Amplify’?

what does Amplication DB mean and what does New Peak Amplitude mean

Decibels are a relative measure of amplitude or energy. With audio dB is related to loudness.

The reference for digital audio is 0dBFs (zero decibels full scale) and that’s the “digital maximum” and digital dB values are normally negative. Internally, it’s as high as you can “count” with 16-bits or 24-bits, etc. (Everything is automatically scaled when you record or play-back, or when you load-into Audacity, so a 24-bit file is not louder than an 8-bit file.)

If your current peak is -6dB and you amplify by +3dB your new peak is -3dB.

Audacity has pre-scanned your file and Amplify will default to whatever gain (or attenuation) you need for maximized/normalized 0dB peaks. For example, if Amplify defaults to +6dB, your current peak is -6dB.

I don’t know what ReplayGain is using for it’s reference but it’s some kind of “loudness” measure and it’s NOT the peak dB level.

More later…

Audacity uses floating-point internally so it can go over 0dB, but regular WAV files, CDs, analog-to-digital-converters (recording) and digital-to-analog converters (playback) are all hard-limited and you’ll get [u]clipping[/u] (distortion) if you try to go over so 0dBFS is considered the “digital maximum”.

Peak levels do not correlate well with perceived loudness. Sometimes RMS dB level (a kind of average) is used to approximate “loudness”, but ReplayGain uses a more advanced algorithm. There is an RMS spec for audiobooks. And in the case of audiobooks, everything is spoken word so matching RMS levels works well to match volumes.

A dB change will affect all of the dB measurements/calculations by the same amount. If you decrease the volume by 6dB, the peaks will go down by 6dB, the RMS (and average) will go down by 6dB and the acoustic loudness will go down by 6dB, etc.

The reference for acoustic loudness of 0dB SPL (sound pressure level) is approximately the quietest sound that humans can hear so SPL levels are normally positive. There is usually no calibration between digital levels and acoustic levels but there is a direct correlation. (If you reduce the digital level by 6dB the acoustic loudness will also go down by 6dB. Of course, that assumes you don’t touch the volume control or get closer to the speakers, etc.).

Perceived loudness is related to the short-term average [u]weighted[/u] to account for the fact that our hearing is most-sensitive at mid-frequencies, and [u]adjusted[/u] for the ear’s varying frequency sensitivity at different volumes.