I love RX by iZotope. I especially love the dialogue/vocal de-noise module. I just click the button, sample a bit of the background noise, and run the algorithm. Next thing you know the noise is gone and there is a minimal effect to the dialogue or vocal. The one downfall of RX is its price. It is completely worth it if you need to use it a lot, but for those of us on a budget the price tag can be a bit hefty… especially if we only need it every now-and-then.
This got me thinking. Is there a way to get similar results using audio effects native to Ableton Live?
And, as it turns out, there is. It isn’t on the same level as RX, but it gets the job done and what’s more, it is automated. Let’s find out how to do it!
First record some dialogue or a vocal. I am using this clip I recorded directly into my phone. This is what the original sounds like +9.5 dB.
AUDIO SAMPLE: original plus 9.5 dB.wav:
Now, before we get into the de-noising process we can clean this up a little bit. I am going to throw a simple EQ on it to boost the high frequencies and trim the ultralow end.
AUDIO SAMPLE: original plus 9.5 dB - EQ.wav:
Now for the important part. This needs to be done before we being the de-noising process.
We need to use a compressor to even out the dynamic range of the vocal. Don’t worry though, it’s really easy to do. Just put a Glue Compressor after the EQ and turn the threshold down to -15 dB.
This value is going to be different for every vocal clip.
What you are looking to do is bring those really high peaks down with some gain reduction so the overall volume is closer to being the same level throughout the clip. This is going to help us in the near future. Obviously you don’t want it to be flat, but bringing the peaks down a couple dB should be fine. You can see how many dB of gain reduction is being applied by watching the Glue Compressor’s needle and how far it peaks.
AUDIO SAMPLE: original plus 9.5 dB – EQ - Glue.wav:
It may be a little difficult to hear the difference so here are what the before and after waveforms look like.
Now we are ready to start the de-noise process. What I do is duplicate the original channel, so we can always go back. Then on the new duplicate I freeze and flatten the track making the EQ and Compression permanent.
Next, I duplicate that new channel. That should leave you with 3 channels in total. You can deactivate the audio on the first channel as we don’t want or need to hear it.
Now take a Utility Audio Effect and drop it on the third channel. Then click the Phz-L and Phz-R buttons at the bottom of the device. This is inverting the waveforms for the left and the right channels. If you play the second and third audio tracks together right now you should hear nothing as the exact opposite waveforms will cancel each other out when they get mixed together in the master channel. Don’t worry, that’s OK!
Now, for the magic.
Solo the third channel—the channel with the Utility device on it. Take the Gate Audio Effect and drop it behind the Utility device. Click the “flip” button, because we want the gate to allow everything except what passes the threshold. Also turn on the “Lookahead” to 10 ms - I found this helps with the very short attack settings we will be using.
Now you need to adjust the Threshold and the Return parameters so the vocal is set right in between the two.
When the dialogue is being played you should hear the noise section but the vocal itself begin to be cut out. Get it close. Don’t forget to set the floor to, so that you can’t hear any of the dialogue or vocal coming through. We want just noise.
This is the goal here. You want the gate to cut out the audio when there is someone talking or singing and ONLY at those moments. The Attack and Release parameters will help to smooth things out after the other parameters are set.
This were my final Gate settings.
Now we can play the second and third channels together and the large areas with just noise will be completely quiet and the dialogue will be allowed to pass through. That will save a boat load of time right off the bat!
Now, the part with the dialogue with still have the noise, so this method is a little limited. But, overall we already have a massively improved result that will handle itself no matter how long to the audio clip is.
However, there is one more trick we can pull off to clean up that final output a bit more.
Go back to the original clip, drop a spectrum analyzer on the channel, and loop a section of just noise. Look for any clear peaks in the spectrum. That is where the noisiest parts are. As you can see from the picture below there is a big problem around 15 kHz and something fishy around 330 Hz.
We can use an EQ 8 on the master channel to precisely remove those sections. Now when the dialogue comes through the gate the noise will be reduced a bit at the frequencies. Here is what my master channel EQ looks like.
Now all we need to do is take a listen to the final output. And, to be honest, I am rather impressed with the results. Go ahead and A|B the original with the final and let us know what you think!
AUDIO SAMPLE: final output.wav:
AUDIO SAMPLE: original plus 9.5 dB – EQ - Glue.wav
Learn more Ableton Live tips and tricks in the AskAudio Academy here.