"Smart" normalization / compression plugin? Or what do broadcasters / streamers use?

[wwenze]

One part of processing audio tracks for my own consumption is to reduce the number of cases where I get jumpscared because of a sudden loud increase in volume for some reason.

Currently I select the offending duration and then manually de-amplify by an arbitrary gain. Example below:

(Left is dB scale)

Is there a better way to do this? Like some form of smart compression plugin that analyzes the volume of surrounding region and acts when there is a sudden change over a period?

Or just any quick method to retain average volume while reducing dynamic range with the effectiveness not affected by the average volume?

Even better if it can use average power correlation instead of peak value since the former is what really matters yet it is hard to see using waveform.

 

[fieldcar]

I use stereo tool. You can try the trial and see how it works with a beep every few minutes. The feature you're looking for is AGC (automatic gain control)

Stereo Tool

Stereo Tool is an audio processor for FM, AM, DAB, TV, streaming and home use for Windows, Linux and Mac OS X

www.stereotool.com

 

[Matthias McCready]

If you are processing tracks you will probably want to put a limiter somewhere in that process.

Overall some compression here is your friend; yes, we all know "too much can be bad" and can ruin tracks, however an appropriate amount is useful.

 

[Tupisac]

I think most streamers use limiters and compressors built in OBS:

 

[DVDdoug]

Or, what do broadcasters / streamers use

Broadcasters use a combination of compression and limiting.

Limiting is required because it's illegal to over-modulate.

Most streaming services don't compress or limit. They pre-scan the file and adjust it up or down (before the song/program starts) to hit an LUFS target. But they won't boost it into clipping so it can end-up quieter than the target.

...ReplayGain works similarly but ReplayGain was created before LUFS was standardized so it uses a different algorithm and it's scaled to dB SPL (acoustic) instead of DBFS (digital). An SPL reference is used because of the Equal Loudness Contours. LUFS probably assumes an SPL level too but I haven't studied it that deeply.

 

[ZolaIII]

What is used widely currently but mostly in modified form (in the range of - 14~-18 LUFS, recently mostly - 18 but if you have strong enough amplifier it's recommended to stay at standard - 23) from streaming services and broadcaster's is EBU R128. In EU it's even legal obligation.

EBU R 128 - Wikipedia

en.m.wikipedia.org

As it's a open source standard you can find plugin's for it (if not already integrated) for most popular software player's.

 

[weesch]

Hi
i have trying many plugins compressor.
and this one take my attention

UltraChannel

UltraChannel The Ultimate Channel Strip Professional Audio Engineering For All UltraChannel™ is a high-precision channel strip that includes a routable Gate, Parametric EQ, two Compressors, […]

www.eventideaudio.com

i mastering this song with cubase and ultrachannel

Last Days Saints | weesch

Last Days Saints by weesch, Rock music from Paris, IDF, FR on ReverbNation

www.reverbnation.com

eventide O-Pressor work great .
cheers

 

[dasdoing]

what you need is a compressor/limiter reacting to LUFS meassurements. you want it to react to momentary and/or short term loudness at a given threshold.
it should be possible with the leveler of this compressor: https://www.soundradix.com/products/powair/
you would set the target level at a momentratry and/or short-term LUFS level that you feel unconfortable. by setting the noise-floor level at the same level as the target you garantee that nothing will be made louder at all.
granted I never tried this out, but it looks like the tool should achieve this.
this would be the relevant section in their presentation video:

I imagine that there are broadcasting tools doing similar, but are probably much more complex and more expensive

 

[kemmler3D]

This is a classic use case for any dynamic compressor. There are plenty of free VST options that will do the job. Avoid "classic" or "vintage" or "analog" stuff because it will tend to add distortion intentionally. Here's one that should work: https://www.kvraudio.com/product/pajsoil-eis-by-pajczur but there are dozens out there.

Setting up a compressor is not intuitive at first. Hopefully I can give some guide to use. In general, for this use case, you want:

Somewhat high threshold - (maybe -6dB?) The threshold is the level at which you want the compressor to start turning the volume down. Below that level, it leaves the signal alone. You will set this to the level in dB below full scale that you feel is loud enough, and higher is too loud.

Fast attack (0.0ms probably) - this indicates how long it takes the compressor to turn the volume down, after the signal goes over the specified level.

Slow release (>1000ms probably) - this indicates how long the compressor takes to turn the volume back up. You want this to be rather slow for playback, since if the compressor "turns on and off" quickly, it creates audible "pumping".

Ratio: (infinite or at least 1:20) - this is how many decibels the audio is turned down for each decibel over the threshold. A limiter uses infinite ratio (no peaking over the threshold at all) but a high ratio anywhere from 10-30 can sound more natural while still keeping peaks under control.

No pre or makeup gain. 0.0dB (you only want to turn the signal down, not raise the overall level to reduce dynamic range.)

Soft knee (the transition between gain reduction and normal operation) - hard knee or limiting goes from 0% reduction to 100% as soon as the level crosses the threshold. Softer knee (smoother transition) will tend to reduce the audibility of the compressor activating.

 

[dasdoing]

This is a classic use case for any dynamic compressor. There are plenty of free VST options that will do the job. Avoid "classic" or "vintage" or "analog" stuff because it will tend to add distortion intentionally. Here's one that should work: https://www.kvraudio.com/product/pajsoil-eis-by-pajczur but there are dozens out there.

Setting up a compressor is not intuitive at first. Hopefully I can give some guide to use. In general, for this use case, you want:

Somewhat high threshold - (maybe -6dB?) The threshold is the level at which you want the compressor to start turning the volume down. Below that level, it leaves the signal alone. You will set this to the level in dB below full scale that you feel is loud enough, and higher is too loud.

Fast attack (0.0ms probably) - this indicates how long it takes the compressor to turn the volume down, after the signal goes over the specified level.

Slow release (>1000ms probably) - this indicates how long the compressor takes to turn the volume back up. You want this to be rather slow for playback, since if the compressor "turns on and off" quickly, it creates audible "pumping".

Ratio: (infinite or at least 1:20) - this is how many decibels the audio is turned down for each decibel over the threshold. A limiter uses infinite ratio (no peaking over the threshold at all) but a high ratio anywhere from 10-30 can sound more natural while still keeping peaks under control.

No pre or makeup gain. 0.0dB (you only want to turn the signal down, not raise the overall level to reduce dynamic range.)

Soft knee (the transition between gain reduction and normal operation) - hard knee or limiting goes from 0% reduction to 100% as soon as the level crosses the threshold. Softer knee (smoother transition) will tend to reduce the audibility of the compressor activating.

using a traditional compressor will mess up with all the peaks. it's not the peaks that bother him, but average loudness level. he needs a compressor that reacts to loudness, may it be RMS or LUFS. LUFS is much more reliable.
With the compressor I sugested above he could, once ajusted, theoreticly blindly throw his whole collection threw it and it will only act on the probably rather rare parts he finds anoying

 

[kemmler3D]

using a traditional compressor will mess up with all the peaks. it's not the peaks that bother him, but average loudness level. he needs a compressor that reacts to loudness, may it be RMS or LUFS. LUFS is much more reliable.
With the compressor I sugested above he could, once ajusted, theoreticly blindly throw his whole collection threw it and it will only act on the probably rather rare parts he finds anoying

Yeah, good point. It is not clear from the original post how long the sections of too-loud audio are. I guess for sections lasting only 1-3 seconds, a traditional compressor is appropriate. If we are talking full songs, an RMS-aware AGC tool would work, like you said, or just replaygain or something.