Skip to main content
Participating Frequently
March 16, 2017
Answered

Why does my voice sound like I'm talking underwater?

  • March 16, 2017
  • 1 reply
  • 12203 views

Hello,

How do I stop my voice sounding like I'm talking under the water whenever I remove the background hiss from my file?

Thanks

This topic has been closed for replies.
Correct answer Bob Howes

Both the previous answers are correct.  The best place to get rid of hiss is in the original recording--if there's a continuous hiss, there's a problem with your recording system.  It's worth doing some detective work to see if you can track it down.

But of course that doesn't help you with things you've already recorded.  For those, Steve is right.  To go into a bit more detail:  instead of just doing one pass of NR at the default settings, you'll get far better off if you do 3-5 passes with very light compression, raising the FFT setting between each pass.

Basically, this means selecting a noise print in the normal way, then run a pass of NR at the default FFT size, probably 2048 on your programme.  The FFT size is on the Advanced menu--you'll have to click on the advanced tab.  When you run this pass, set the NR slider to around 10 or 15% and the Amount to about 10dB.  Select all and do a pass.

Then raise the FFT size to 4096, grab another noise print and run another pass with the sliders at the same setting.  Keep doing this, advancing the FFT each time until you're either happy with the sound or you run out of FFT changes.  One thing to remember is that the higher the FFT size, the higher the frequency it can eliminate.  With a bit of experimenting, you may be able to cut the number of passes by using the FFT sizes right for your hiss.  FYI, you can find a detailed tutorial on this on the "Audition Learn" link under the help menu.

...but do try to get rid of the hiss at source for your next recording.

1 reply

SteveG_AudioMasters_
Community Expert
Community Expert
March 16, 2017

Don't take so much off!

Do several passes at different FFT settings, and take less off at lower ones. If it's mainly hiss you want to get rid of, then the higher FFT settings are the ones to go for - but don't overdo it; that's when you get the strange watery noises.

ryclark
Participating Frequently
March 17, 2017

Better still try and reduce the noise at source before it gets recorded. Where is the noise coming from? A decent mic and audio interface shouldn't produce any noticeable noise when recording.

Bob Howes
Bob HowesCorrect answer
Inspiring
March 17, 2017

Both the previous answers are correct.  The best place to get rid of hiss is in the original recording--if there's a continuous hiss, there's a problem with your recording system.  It's worth doing some detective work to see if you can track it down.

But of course that doesn't help you with things you've already recorded.  For those, Steve is right.  To go into a bit more detail:  instead of just doing one pass of NR at the default settings, you'll get far better off if you do 3-5 passes with very light compression, raising the FFT setting between each pass.

Basically, this means selecting a noise print in the normal way, then run a pass of NR at the default FFT size, probably 2048 on your programme.  The FFT size is on the Advanced menu--you'll have to click on the advanced tab.  When you run this pass, set the NR slider to around 10 or 15% and the Amount to about 10dB.  Select all and do a pass.

Then raise the FFT size to 4096, grab another noise print and run another pass with the sliders at the same setting.  Keep doing this, advancing the FFT each time until you're either happy with the sound or you run out of FFT changes.  One thing to remember is that the higher the FFT size, the higher the frequency it can eliminate.  With a bit of experimenting, you may be able to cut the number of passes by using the FFT sizes right for your hiss.  FYI, you can find a detailed tutorial on this on the "Audition Learn" link under the help menu.

...but do try to get rid of the hiss at source for your next recording.