Re: Help with Vocals
Posted: Sat Oct 23, 2010 12:45 am
I was working on a similar issue recently and I was able to make a "noticeable improvement" with equalization and a "vocal exciter". (It still doesn't sound like a modern, clean, digital recording.)
As Koz said, instrument & vocal ranges overlap (and of course an instrument can be playing the exact same note that the singer is singing), but I was still able to "bring out the vocals" to some extent. Every recording is different so you'll have to experiment. (I experimented so much that I don't remember the exact settings I finally chose.) I think I boosted between 300 & 800Hz by about 6dB (with a parametric EQ). I was surprised how much EQ it required! I did write-down the settings that I used... Somewhere.
The "exciter" effect adds upper-harmonics to bring-out "T" & "S" sounds... and cymbals... and distortion... This took quite a bit of trial & error too. Where the "T" & "S" sounds are totally burried or missing, this doesn't help much. Of course, some high-frequency EQ boost can help here too.
I didn't do this, but some (dynamic) compression is worth a try too. Compression makes "loud parts quieter" or "quiet parts louder" (It's mostly used to tame peaks & make "everything LOUD".) You can't reverse things and make the vocals louder than the instuments, but you may be able to make the sounds somewhat more-equal.
I tried something else that didn't work.
Since some male voices were very "thin", tried to make a sub-harmonic exciter effect... I used a bandpass filter and made a copy that only contained the "lower voice" frequencies. I pitch-shifted that down one octive and mixed it back in. It sounded lousy. (I have done something similar to make deep bass notes, and that did work!)
A couple of times, I've been temped to record my own voice and mix-in my "T" & "S" sounds... But I never have.
Any processing that boosts the level (especially things like 6dB EQ boost) can drive the levels into clipping (distortion). So, it's a good idea to normalize (bring the peaks back down to 0dB) before rendering the final processed file. (Most audio editors use 32-bit floating-point for processing & temporary storage so you can go over 0dB without clipping. But, "regular" integer WAV files have a hard-upper limit and you can get clipping whan you save/render to WAV.)
) I always say, "you can't un-bake a cake or un-fry an egg, and you can't un-mix a recording." But, there is a company called UnMixingStation that claims they can do it... for a fee...
As Koz said, instrument & vocal ranges overlap (and of course an instrument can be playing the exact same note that the singer is singing), but I was still able to "bring out the vocals" to some extent. Every recording is different so you'll have to experiment. (I experimented so much that I don't remember the exact settings I finally chose.) I think I boosted between 300 & 800Hz by about 6dB (with a parametric EQ). I was surprised how much EQ it required! I did write-down the settings that I used... Somewhere.
The "exciter" effect adds upper-harmonics to bring-out "T" & "S" sounds... and cymbals... and distortion... This took quite a bit of trial & error too. Where the "T" & "S" sounds are totally burried or missing, this doesn't help much. Of course, some high-frequency EQ boost can help here too.
I didn't do this, but some (dynamic) compression is worth a try too. Compression makes "loud parts quieter" or "quiet parts louder" (It's mostly used to tame peaks & make "everything LOUD".) You can't reverse things and make the vocals louder than the instuments, but you may be able to make the sounds somewhat more-equal.
I tried something else that didn't work.
Since some male voices were very "thin", tried to make a sub-harmonic exciter effect... I used a bandpass filter and made a copy that only contained the "lower voice" frequencies. I pitch-shifted that down one octive and mixed it back in. It sounded lousy. (I have done something similar to make deep bass notes, and that did work!)A couple of times, I've been temped to record my own voice and mix-in my "T" & "S" sounds... But I never have.
Any processing that boosts the level (especially things like 6dB EQ boost) can drive the levels into clipping (distortion). So, it's a good idea to normalize (bring the peaks back down to 0dB) before rendering the final processed file. (Most audio editors use 32-bit floating-point for processing & temporary storage so you can go over 0dB without clipping. But, "regular" integer WAV files have a hard-upper limit and you can get clipping whan you save/render to WAV.)
Our silly joke is you're trying to carefully remove the chocolate from a chocolate cake.
) I always say, "you can't un-bake a cake or un-fry an egg, and you can't un-mix a recording." But, there is a company called UnMixingStation that claims they can do it... for a fee...