Improvement of Auto Duck effect

Recently I used Audo Duck effect by Markus Meyer with a long sound track (audio track from a movie) and voice track (recorded spoken movie subtitles) to mix them both together. The effect was supposed to decrease volume of parts of sound track where voices were about to be mixed in. Most of the result was all right but some voices were not possible to understand because of the background sound still too loud.
Auto Duck effect reduces volume (amplitude) of audio track by choosen amount of decibels instead of reducing volume to a certain common level lower than voice segments if appropriate sound segments are too loud.
I would like to add another options in Auto Duck settings window:

  • New Peak Amplitude (like the one in Amplify effect by Dominic Mazzoni) for particular sound segments;
  • probably even better would be New Average Amplitude.
    So there could be 3 options to choose from or only 1 option but the proper one. As I am not audio expert, I do not know how exactly to achieve the same volume level for all “noisy” sounds in the background of voices. Please check my ideas and help me to improve AutoDuck effect.

That’s an interesting idea mederi and I can see how it would be useful for how you are using the Auto Duck effect.

Usually the “Duck Amount” setting works well as the effect is most commonly used for “radio show announcements” and similar applications where the music track has a relatively constant level. Generally the music would be amplified / normalised to a fairly high peak level before using the Auto Duck effect. With the default settings the “ducked” level would then be below -20 dBFS, which would be sufficiently quiet for announcements to be heard. However, I can see that with a film sound track, using -20 dB Duck amount would be excessive for much of the track as low level sections would become virtually silent, but if you use a smaller amount of “duck”, then it will not be enough for loud sections of the film soundtrack.

There is a relatively easy workaround, as you could apply the Auto Duck effect multiple times to the parts that need greater level reduction.
This is obviously less convenient than being able to do it automatically, but it saves making the effect more complicated.

The problem that I can see with your proposal is that during a “ducked” section (while you are talking), the level of the soundtrack could vary dramatically.
For example, what would you want the effect to do if you were speaking during the selected (shaded) area of this track?
If you have a “target level” of say -20 dBFS and the highest peak (toward the right end of the selection) were adjusted to the target level, then the quiet part will be virtually silent, but on the other hand if the majority of the track were adjusted to -20 dBFS, then the loud part at the right end of the selection would be too loud.
Ideally what you want is a “compression” effect with a fast attack and a fairly slow release so that the “ducked” audio is dynamically adjusted to a target level, but that would make the effect considerably more complex from both programming and user perspectives.

Some sophisticated algorithm is needed to think out for the task of reducing of noisy sounds only to a certain volume level. I would like to ask Markus Meyer, the creator of Auto Duck effect. Perhaps he would be able to improve it. Is it possible to contact him? I am really interested in his opinion, whether it is possible to do something with it.
Then I was thinking of some volume map (no change / increase / decrease) for the audio track. It could be generated faster than making changes directly in audio track. Then exporting of audio track could consider the volume map into a result. Does it make any sense? I have really no further details, just some “crazy” idea again :slight_smile:

I don’t think that Markus is currently active in Audacity development so I don’t know how you would contact him.
Are you aware of the “Envelope Tool”?
For now at least you could use the Auto Duck effect to make most of the changes (processing in sections if necessary), then use the Envelope Tool to fine tune any awkward sections.

Yes, Steve, you are right. The audio track’s Envelope is some kind of “volume map”. Would it not be faster if effects like Auto Duck adjusted the envelope instead of making direct changes in the track? Because the effect is quite slow and in combination with slow computer like my one it is really slow :slight_smile:

I’m not sure if it would be faster because I don’t know how much of the time is spent applying the volume changes and how much time is spent analysing the audio and calculating what changes need to be made. I suspect that most of the time is in the analysis stage. However I still think that it is an excellent idea as volume envelopes are non-destructve (they don’t change the actual data until the track is “rendered” or “mixed”) and the volume envelopes are adjustable.

If you are an experienced programmer you might like to try developing this? If not, you might like to think through exactly how you would like it to work and post it as a proposal in this forum thread. There’s no guarantees if and when proposals are implemented, but proposals from this forum board are eventually transferred to the “Feature Requests” section of the Audacity wiki Missing features - Audacity Support Some of the developers monitor the wiki band it gives them an idea of the features that users would like to see. A clear proposal here would make it easier for Peter who transfers proposals from here to the wiki to put the idea across clearly. This idea gets my vote :slight_smile: