Increase volume only of of low-volume parts of the audio, while keeping high-volume parts as-is

I’m having audio files which have majority of audio at normal listenable volume, interspersed with many almost-unhearable low-volume parts (think presentation recording, where only presenter has a mic, and questions from audience are picked up by it, but at much lower volume, obviously).
Turning the volume up and down in audio player all the time works but is highly annoying and does hearing-damaging extreme loudness at parts so I’m looking to preprocess audio file so it is listenable normally without user interaction.

Analyze / Measure RMS says the normal parts are at -30dB, and low-volume parts are at -65dB. I’d like to increase low-volume parts volume close to volume of “normal” parts, in as easy/automated way as possible.

Currently, I found following workflow which produces quite listenable results:

  • select all audio, Analyze / Label sounds, select -50dB and Average level, type Region between sounds - which correctly marks low-audio parts
  • doubleclick on first low-volume label to select it, and choose Effects / Normalize with peak to -10dB- which correctly increases its volume to be on-par with “normal” audio.
  • repeat previous step for each and every label

Of course, while the result is what I want, the workflow is very slow, manual and tedious. I’d love to have more of “one-click” solution. Is it possible to automatically repeat the for every label? Or any other suggestions to accomplish what I want?

I’ve found mentions of “Automatic Gain Control” (which I’m unable to find?) and “Compressor” (which seems to make situation only somewhat better than bad original, and nowhere near as good as my workflow, and also seems to distort the “normal” volume which I’d prefer to leave intact). I’ve also tried to play with “amplify” and “auto duck” with no luck.

I’m using Audacity 3.2.4 at Debian Bookworm GNU/Linux. I have little to no experience in audio.

LevelSpeech2.ny plugin worth a try: it’s a compressor-limiter combo.
It’s an Audacity-specific plugin, so should work for Linux users.

1 Like

There is probably no automatic process/effect that will match what you’re doing manually.

Dynamic compression reduces the dynamic range (or “dynamic contrast”) by making the loud parts quieter and/or the quiet parts louder. Most compressors “push-down” the loud parts, and in most cases “make-up gain” is applied so you end-up with “everything louder”.

Limiting is a kind of fast compression and automatic volume control (or automatic gain control) is a slow-kind of compression.

You can try Chris’s Compressor. It may be what you’re looking for.

Compression is one of the “big 3” effects (along with equalization and reverb) and there are LOTS of 3rd-party compressor plug-ins that may, or may not, work in Audacity.

There is a Leveller Effect “hidden” as a “distortion” effect. (It shouldn’t sound like distortion).

GoldWave ($60 after free trial) has a compressor/expander with some interesting/unusual options. There is a “boost quiet parts” preset. GoldWave confuses the definitions of compression & expansion but the preset descriptions are accurate so you can start with a preset and then experiment with the settings. (I have the Windows version of GoldWave and I have not tried the universal online “Infinity” version.)

1 Like

Thanks, by description/example in that thread it sounded exactly like what I needed. It installs fine (Tools/ Nyquist Plugin installed followed by Tools/ Plugin manager / Rescan) and shows under Effects / Level Speech 2, but I’m unable to find parameters that work for my use case. It seems I need to lower limiter threshold all the way down to 0.1 so it will amplify low volumes, but then it also have noticeable distortions at normal audio :man_shrugging:

Ah, that’s too bad. I was hoping one could automate that with a macro. I guess .ny plugin might be able to iterate over labels and apply effect to each of them, but that sounds like waaay too steep learning curve. Unless someone has simple example that iterates over labels and does something with audio?

Hmmm, I’ve tried playing with it, but unless I crank up degree of leveling up to 5, it does not increase the volume enough, and if I do, it does distort the normal parts too :cry: . Quick playing with other sliders doesn’t seem to influence result much.

Thanks, I’ve played with that and it seems interesting. Setting compression level to 0.95 does amplify low volume parts almost as much as I would like it, but it also affects the normal volume. But there are so many options in that one that after an hour playing with it I did not reach ideal results, so I’ve given up.

I’ve found other related threads: How to increase volume of portion of audio track? and Raise volume of lower parts without affecting the rest

In the end, I’ve found that running included Compressor with -50dB threshold, -75dB noise floor, 10:1 ratio and make up gain for 0dB checked, produces acceptable results without too noticeable distortion on “normal” parts (might need to run it twice on whole audio for exceptionally big volume differences, or audio tracks that has multiple different levels of “low volume”).

While I’d still prefer a way to automate my workflow as it leaves “normal” audio parts 100% untouched, unless someone has simple example how to iterate over labels, I’ll live with this solution.

Levelator is a free multi-platform standalone program

It has no controls, (you could apply it twice if necessary).