Fixing variable sound level (DJ Mix)

[Xyz_39808]

Can't you just use the Envelope tool as you would the master volume knob on your mixer? Though, instead of avoiding red lights, you'd be, in essence, normalizing the volumes.

But please, do keep in mind that differences in dynamics are important.

[DJCraft]

But please, do keep in mind that differences in dynamics are important.

How do you mean exactly?

[kozikowski]

I'm going to point you to another thread that deals with this...

http://audacityteam.org/forum/viewtopic.php?f=15&t=8116

I posted this for a completely different reason, but I got to mess with this tool last night and it did a very remarkable job. I used essentially the default numbers and I'm planning on using the tool like that this weekend.

If you follow the developer's narration, he was making a lot of the same program performance decisions that you are asking about. It's extraordinarily difficult to get a compressor to do this job without seeming to be doing anything at all.

One of his notes had to do with the older existing compressor tools in Audacity. He didn't like them very much because they made many wrong philosophy decisions. That's what got my attention. I didn't like them, either, and for those exact reasons.

Koz

[Carol752]

With a DJ mix where some songs are louder than others and also some songs that are too loud but adjusted (during recording), what would be the best way to even this out?

Full Project


I've tried zooming in and normalizing (no dc offset, -6 dB sometimes) as picture below, but I can spend hours on this but there's not much effect on an 80 minute project.

Pre-Normalized


Normalized


I've thought about normalizing to +6 to get rid of the peaks but that sounds like it will just cut the peaks out and make them clip.

[DJCraft]

Carol752, if you wrote anything I missed it. I only see a quote :/

I gave the compressed version another listen in the car and noticed that during parts of the song where the bassline drops out all of the high end gets really loud (ambient synths etc). So the compressor, with the default settings, isn't going to work. Does anyone know if there are alternate settings that would prevent that?

Also, is there anything like the envelope tool, but automated for bringing down the louder parts (say anything that 2x or 1.5x louder than the everything else around it)?

[DJCraft]

I tried out the envelope tool and the result of 5minutes or so of work is about half a second worth of audio:




Zooming out you can quickly multiply the amount of work it would take to get just 5 seconds done:



Are there any plugins that can do this automatically?

[steve]

I gave the compressed version another listen in the car and noticed that during parts of the song where the bassline drops out all of the high end gets really loud (ambient synths etc).

I think what you need is a multi-band compressor. I don't know of one for Audacity, but it's easy enough to reproduce the effect, and although it is a bit more work than just running a compressor at its default settings, it is considerably quicker than manually entering thousands of envelope points.

To produce a simple multi-band dynamic compressor effect:

In this case we will just use two bands, bass and treble, though the method can be expanded to more bands if you wish.

1) Open Audacity and Import your audio track.
2) Select the entire track and duplicate it (Ctrl+D)
3) Use the highpass filter set to your chosen cross-over point (say 800Hz) and a roll-off frequency set to 3B per octave (this is actually 6dB per octave in current Audacity versions).
4) Use the lowpass filter on the other track with the settings identical to what you used in step 3.
5) Apply dynamic compression (use your favourite compressor with whatever settings you require) to each track individually.
6) Listen to both tracks together, and if you like it, Mix and Render. If it's not quite right, use "Undo" and go back to step 3.

The advantages of multi-band compression are that you can apply different compression amounts to different frequency bands. Also, if there is a heavy bass hit, with normal (single band) compression, this will cause all of the audio to dip in level, making a noticeable drop in the upper frequencies. Using multi-band compression, this undesirable side effect is avoided.

[DJCraft]

To produce a simple multi-band dynamic compressor effect:

In this case we will just use two bands, bass and treble, though the method can be expanded to more bands if you wish.

1) Open Audacity and Import your audio track.
2) Select the entire track and duplicate it (Ctrl+D)
3) Use the highpass filter set to your chosen cross-over point (say 800Hz) and a roll-off frequency set to 3B per octave (this is actually 6dB per octave in current Audacity versions).
4) Use the lowpass filter on the other track with the settings identical to what you used in step 3.
5) Apply dynamic compression (use your favourite compressor with whatever settings you require) to each track individually.
6) Listen to both tracks together, and if you like it, Mix and Render. If it's not quite right, use "Undo" and go back to step 3.

The advantages of multi-band compression are that you can apply different compression amounts to different frequency bands. Also, if there is a heavy bass hit, with normal (single band) compression, this will cause all of the audio to dip in level, making a noticeable drop in the upper frequencies. Using multi-band compression, this undesirable side effect is avoided.

Nice one, I'll give this a shot. is 800Hz the recommended frequency for splitting the two bands? Is there a benefit to doing 3 or 4 bands as opposed to 2? If so what frequencys do you suggest splitting at?

[steve]

is 800Hz the recommended frequency for splitting the two bands? Is there a benefit to doing 3 or 4 bands as opposed to 2? If so what frequencys do you suggest splitting at?

It all depends on the music.

It's best to try and keep the crossover frequencies out of the most critical frequency bands otherwise you can get strange phasing effects. Also, if you use more bands, then you will need to use sharper cut off frequencies in the high/low pass filters, but that can cause resonance to occur close to the cut off frequency.

4 frequency bands is likely to be the maximum that is useful, but often a simple 2 band will provide just as good results with a lot less effort.

A good compromise would probably usually be 3 bands, somewhere around 20Hz to 400Hz, 400Hz to 3kHz, 3kHz to 20kHz. If you have,say, a synth bass track that is pumping out a lot at around 300 Hz, then you might find it better to push the lower crossover frequency up a bit, say to around 600Hz, or if you have a lot of deep bass below 100Hz, then you may find that a simple 2 band split at around 300Hz with a sharp cut-off does the best job.

Practice on short samples and keep a note of your settings, after a while you will probably find that you have certain favourite settings.

Remember that while you have the frequency bands split you can also adjust the levels of each band.

[kozikowski]

This is getting seriously surreal. Is there any reason we're not using the multi-band compressor built into Audacity 1.3.5?

Effect > Apple: AUMultibandCompressor

That tool can have serious musical mix problems, but it does smash everything down to a straight waveform on the TimeLine. More importantly, did you try that custom compressor tool I mentioned in an above post. If all you want to do is retain the character, flavor, texture, and presentation of the original performance without the loudness variations, that's exactly what that tool does.

http://audacityteam.org/forum/viewtopic.php?f=15&t=8116

Pull down the two sound clips and compare them. I know it's not screaming rock or dance music, but work with me here. The abrasively loud segments are pushed down and the orchestration is pushed up. You can listen to that whole hour-long show in the car without touching the volume control.

Once you get it that far, you can apply more effects, but if you're mixing dance music, you don't want to take any more moxie out of the performance than you have to. I don't like dance music that's been sterilized.

You can continue to mess with the performance waveform by waveform, but you'll continue to have musical problems if you're zoomed in that tight. Loudness and volume happen in a much broader and sloppier way.

Do you have your "before" performance posted somewhere? Try not to digitally compress it. That can cause problems with sound experiments.

Koz