Hello!
I received a batch of .wav files to mix for an album. The artist has a huge following but literally records in his bedroom (honk).

The microphone he used to record the vocals has an overwhelming amount of high-end above 10 kHz (I’d say 12–17 kHz). On some tracks, however, he decided to apply a high-cut filter to the vocals, so the voice blends better into the mix.

On the tracks, if I apply a 12dB roll-off at 10kHz, the vocals sound better but the track loses its brilliance; if I use a match-EQ, multiband compression, a de-esser, or a dynamic equalizer just on this range, I still can’t fix the fact that there’s this “harsh vocal noise.”


Soothe is probably the best solution, but I’d also like to learn some non-automated methods or if you have any secret tips for these situations—I imagine you’ve encountered a harsh vocal in a mix quite a few times

 

Try (on macos) Macxvideo AI. I find the VoiceFilter v3.1 very good. You can isolate that frequency and beet them with a dynamic EQ. If one EQ point is not sufficient 2 with different aperture - maybe 1@ 2 and the second one @ 4 aperture could make the trick.
More opportune and specific choices you could find with IZotope RX if Soothe didn't help.

 

I wanna bring to your attention a Zynaptique plug called UNFILTER. I managed to rescue quite a few bad recordings using this. The plug is a deconvolver, it takes your shitty mic recording and listens to filtering that stays constant (such as that imposed by the mic, or comb filtering in the room etc). Then it creates a detailed linear phase EQ curve to de-filter it for you, with various useful tuning settings so you can work fast and blend the result to taste.

My workflow would be:

- Run the raw shitty vocal through Unfilter to make it sound "decent" if possible
- Put the "linearized" vocal which now sounds unnatural, through a good mic simulation of your choice/taste, so it can be made to sound like he recorded this shit on a Neuman previously owned by Larry King.
- Continue to work with the result using de-essing and dynamic EQ as needed from here.
- Final cherry on top, experiment putting a medium-narrow deep notch at around 5khz with any EQ. This is naturally present in most vocals, but UNFITLER will have artificially ironed it out, so it's worth checking if the extra manual cut actually seems more pleasing.

Usually the outcome is a recording that is not bad, the tradeoff is that it will probably sound a little crunchy, but at least no longer unlistenable due to HF overload and for some uses/mixes the crunch might not be entirely undesirable.

PS: Unfilter is capable even of removing his LPFs that he stuck on some parts of the vocal, if you're desperately needing to fix those too.

 

Can you supply a 2 or 3 second audio example where the harsh vocal noise is prominent? Maybe someone will have success removing it, and then share their process.

 

How to Tame Harsh Vocals

When working with a harsh vocal, you can typically tame it by attenuating frequencies above 2kHz and below 12kHz using a subtle bell filter via an EQ. Attenuating 3-5kHz will help significantly, as will using a de-esser on 5kHz to 12kHz to reduce harsh-sounding sibilance.

Read more: www.sageaudio.com/articles/how-to-tame-harsh-vocals