Tip - Removing resonances with Auto-EQ's

[vectorwarrior]

This is perhaps common knowledge, but I've been using Auto-EQ's for a while and hadn't tried this until today.

Using the 'draw' functionality of the auto-eq's you can compare a section of audio with a flat frequency response to auto-eq out any room resonances or frequency imbalances. Obviously you'll need to compensate for lower high frequency energy by making the curve downward towards high frequencies, but that's easily achieved.

1.jpg

-Open an auto eq
-In Analyse Source, click 'Draw'
-You may need to right click and 'clear points' to reset
-Create a straight curve, I find 0dB on the left side (20hz) and -12dB on the right (20khz) works well
-You may benefit from switching on snapping in right click menu to make this easier
-Now analyse some of your audio with the Analyse Target
-Then Equalize blue to sound like red
-You may need to increase the number of bands in the auto-eq settings window
-You may also probably want to raise the bottom and lower the top frequencies in the same window to tailor the result
-If the issues/resonances are very tightly focused you'll probably want to reduce the 'smoothness' amount in the main window to 1% or whatever
-Finally, you may want to adjust the wet/dry amount as it'll probably be overkill at 100%

Doing the above, I was able to quickly turn some nasty voice over into something sounding really nice very quickly. I was surprised how well it worked so I thought I'd share. Obviously works better on audio that is supposed to be full frequency or you'll just be boosting noise (again, use those top/bottom frequency limits to tailor that).

I'd also add that I found the 'draw' approach sounds considerably better than using the 'load' to load up pink noise, as I was able to tailor the curve to suit the material and having the curve range from 0 to 'minus-something' resulted in considerably better results for some reason.

Hope someone finds this interesting!

[Tagirijus]

Nice idea. Thanks for sharing!

[MeldaProduction]

[alexis1]

This is perhaps common knowledge, but I've been using Auto-EQ's for a while and hadn't tried this until today.

Using the 'draw' functionality of the auto-eq's you can compare a section of audio with a flat frequency response to auto-eq out any room resonances or frequency imbalances. Obviously you'll need to compensate for lower high frequency energy by making the curve downward towards high frequencies, but that's easily achieved.

1.jpg

-Open an auto eq
-In Analyse Source, click 'Draw'
-You may need to right click and 'clear points' to reset
-Create a straight curve, I find 0dB on the left side (20hz) and -12dB on the right (20khz) works well
-You may benefit from switching on snapping in right click menu to make this easier
-Now analyse some of your audio with the Analyse Target
-Then Equalize blue to sound like red
-You may need to increase the number of bands in the auto-eq settings window
-You may also probably want to raise the bottom and lower the top frequencies in the same window to tailor the result
-If the issues/resonances are very tightly focused you'll probably want to reduce the 'smoothness' amount in the main window to 1% or whatever
-Finally, you may want to adjust the wet/dry amount as it'll probably be overkill at 100%

Doing the above, I was able to quickly turn some nasty voice over into something sounding really nice very quickly. I was surprised how well it worked so I thought I'd share. Obviously works better on audio that is supposed to be full frequency or you'll just be boosting noise (again, use those top/bottom frequency limits to tailor that).

I'd also add that I found the 'draw' approach sounds considerably better than using the 'load' to load up pink noise, as I was able to tailor the curve to suit the material and having the curve range from 0 to 'minus-something' resulted in considerably better results for some reason.

Hope someone finds this interesting!

Hi, thanks for posting this!

I'm having trouble understanding how it works. For example, I don't see anything about making deep narrow cuts at targeted frequencies to remove room resonances, just a "make this voice sound like this curve". If I'm understanding that correctly, won't that tend to smooth out the distinct frequency characteristics that help distinguish voices from each other?

Thanks for explaining!

[vectorwarrior]

Well it's kind of both to be honest, so the approach doesn't work on all material, sometimes you want unusual frequency curves. But if course you have complete control and this is also why the wet dry control is so important (it reduces the curve rather than just mixing the dry in).

The point is that by analysing a long period of target material you get an average of the frequency response. So if there's, say, a nasty lump at 1khz it'll stand out from the curve. Of course, this is where finding the appropriate curve type by drawing is so important, it requires a bit of trial and error. By using the auto eq, the eq will attempt to flatten that lump and you can choose to hone in on the problem area only with the limit controls.

If you want the unusual curve of a recording but want to fix a specific area, you can learn the source and then draw out the problems, although that's the theory, I'm not sure you can actually draw on learned graphs.

[alexis1]

...

The point is that by analysing a long period of target material you get an average of the frequency response. So if there's, say, a nasty lump at 1khz it'll stand out from the curve. ...

Thx, vectorwarrior, this is one of those things I'll have to read a few times.

Conceptually, it's unclear to me how one would know whether the 1kHz hump was a bad room resonance, or a specific identifying characteristic of the vocal one would want to keep.

Thanks again!