How to amplify after noise removal w/out adding noise

Hi, I am a new user and am feeling pretty overwhelmed. I am a court reporter and have received some poor-quality audio, but with tips from a friend, figured out how to use noise removal, which was miraculous, however, now the speaker voices are too low for me to hear. When I amplify, it puts some of the distorted noise back in and I am back to square one, unable to clearly make out what the speakers are saying. Is there anything I can do? I tried normalize, also, and it put the distortion and interference back in, too. My problem might be the values I am using? Please keep in mind I am super new, and although I have read many forums and watched some tutorials, I am still in way over my head, so instruct me like a toddler!

Re: noise removal …

Below is the equalization “curve” for a 300Hz -3000Hz band pass filter

You could try something a bit wider first, e.g. 200Hz - 5000Hz

Thanks, Trebor. I did the equalization to close to the example you had. I tried it both before and after noise removal and alone. It does increase the volume on its own, but once noise removal is added in, the volume is taken away. SO frustrating. I have potential money to be earned just sitting here I can’t do because I can’t figure out how to fix the audio, if it can be fixed.

I’m a bit surprised one of the regulars here hasn’t asked it yet - so I will: How exactly are you doing the noise removal? If the volume goes way down when you do the noise removal, then it sounds to me as if the noise sample includes some of the program that you want to keep. For noise removal to work best you need to sample a little bit of your audio that has noise only, make sure none of the voice that you want to preserve is in the sample. Then with this noise only sample selected, click the “Get Noise Profile” button in the Noise Removal dialog. When that’s done select the whole track, adjust the sliders under “Step 2”, then “OK”. I’ve never used a noise reduction as high as 24dB, normally I set it around 9dB - 12dB. But then again, I’m normally processing audio from a recording of an LP, not a noisy voice track. Some experimenting with noise samples and the control sliders is undoubtedly called for.

I once did make a usable version of a low volume, noisy cassette recording of an interview of my wife’s father. To get anything at all I used noise removal, equalization similar to what Trebor described, and the compressor effect to try even out the volume variations in the voices.


24dB will not be theatrical. The noise will remain inside each word as its spoken, but the spaces between the words will be clear. This sounds really weird and is the edge beyond which you say you have no show and start over.

Yes, the Capture Profile step is critical. Noise Removal will try to subtract everything in the profile from the show. If you select noise only, only the noise will go away. If you capture some of the voice by accident, the voice will go away, too.

This is a Big Problem on a very noisy show. Sometimes, you capture voice with the noise because the noise is so loud you can’t tell the noise isn’t alone. This will really mess up your reduction job and you won’t be able to tell why.

Sometimes, you can select some Noise by itself and apply Amplification before you capture the profile. This gives Noise Reduction more to chew on and sometimes gives you better results. Be sure to UNDO the amplification later.


You could attach a 5-10 seconds of the worst part of the original recording, (before you have attempted to clean it ) to your next post,
and we could tell you if you are flogging a dead horse. Bear in mind anyone can hear and download audio posted in this forum.

Thank you all! PDXRunner, that makes a lot of sense and I will try that immediately to see if that helps. I did not realize I needed to clip a portion without voice on it for the noise profiler to do its job well. I hope I have a portion without voice. Trebor, if I am still floundering, I will post a portion of audio for input from all, but I am embarrassed to say I do not know how to go about that, but before you waste time explaining it to me, I will try the other two suggestions first. I deeply appreciate everyone’s time in posting. I know there are a lot of threads and a lot of people needing help. I just started my own business and feel a lot of pressure to get things right and not lose clients!

See the first link in Trebor’s first post.

Select a bit of a audio (click and drag on a track) then “File menu > Export Selection” and export as a WAV file,
Short WAV files (up to about 6 seconds) can be “attached” to a forum post - look below the compose message box for “Upload Attachment”.

Thank you so much, Steve! And Trebor, I am sorry I didn’t realize I was to open the two links you included above your post in order to help me further.

Here is a sample of how the audio sounds. I am crossing my fingers someone here knows how to fix it, if it can be fixed. The entire piece sounds like this and it’s especially bad when the interviewee is speaking. The room the interview took place is a large, cavernous place, which isn’t helping.

Thank you again for helping me!

In short the patient is terminal.

It sounds like it was recorded via an echoey tube, like an air-vent or drain-pipe.

Then way too much compression has been applied which has introduced the whistley R2-D2 artifacts.

I’ve attempted to recreate this combination of distortions …

To produce this recreation I used a mp3 bit rate of 8Kbps,
for a comprehensible mono voice recording at 16000Hz sample rate you should use a bit rate of not less than 32Kbps, (I’d use 64Kbps).

Once a recording has been damaged by too much compression (too low a bit rate) it cannot be undone.

The detective insists it is not his recorder, but the recordings I have received from him lately, even when not recorded in large, open places like jail bathrooms, sound like this recording. Could it be the way he downloads/saves the audio files?

Thanks for trying to help, Trebor. I was really hopeful. :cry:

Yes, an echo effect can be created by accident when using an audio editor like Audacity, (search this forum for “corridor”).

I’m confident the recording device did not create the whistley R2-D2 compression artifacts, IMO they have been added by using extreme compression settings in an audio editor (like Audacity) to minimize the size of the audio file, e.g. to send the smallest (MP3?) file possible by email, but which has mutilated the recording in the process.

BTW I’ve heard the recording a several times and think I can make out part of it, however to me some of the voices sound like they are speaking with my accent (Scottish) which is a symptom of pareidolia, i.e. my brain may be making this up …

01-02 : “ Huh ?
02-04: Corey “ Aint so crazy about me being in America "
04-05: Scot : “ well I‘ll tell you ”.
05-06: US Lawyer (interrupts Scot) “ it’s not about you ”.
06-07: US Lawyer “ Did you sign it ? ” .
08 : Scot “ Hard Luck ” (sarcastic).
09-10: US Lawyer “ again Did you sign it ? ”.
11: Cockney Lawyer “own up”.
12: US Lawyer “ Correct ”.
12-14: US Lawyer “ they haven’t told you did ? ”.
Cockney Lawyer " No "
13-15: US Lawyer “ OK gotcha ”.
15: US Lawyer “ Corey " (interrupted by Cockney geezer)
15-17: Cockney Lawyer “ the judge will tell you right now ”.

That could be at least part of the problem.
Much of the damage is due to data compression (as Trebor said).
If audio is compressed to a lossy format (such as MP3), some of the sound quality is lost. There is a trade-off with lossy compression of sound quality vs. file size. The loss of sound quality is permanent and unrecoverable. If a file that has been encoded in a lossy format is compressed again (transcoded) to a lossy format (for example, changing a WMA file to MP3), there is further sound quality loss. The sound quality gets worse each time it is encoded.

Ideally the file that is sent to you should be the original recording in whatever format it has been recorded in.

If the echo isn’t an artifact from incorrect use of an audio editor here is a possible explanation …

Thanks Trebor and Steve. I will be relieved and feeling kind of stupid if it is just a matter of the file being compressed for email too many times. That did not occur to me since I have no other problems with other clients who email files to me, but I believe this file had been put on a disk, given to another party, then the file was emailed to me from the disk and third party. I usually receive my files from and not by email and they are normally coming right off the interviewer’s computer after they download the recording, except for this time. I hope the compression is indeed the issue. Thank you for educating me!

Sending an audio file by Email per se does not introduce the compression artifacts.

The file size restriction on email attachments, (5Mb - 25Mb depending on email provider), may have obliged the sender to compress a large audio file, (using something like Audacity), which in this case they have overdone, mangling it in the process.

The reverberating echo and whistley compression artifacts could have been added by the person who created the disk, or the person who copied the disk and emailed you the file.

If the “other party” tried to copy the audio from the disk using the record “what-U-hear” method *, (rather than just copying the audio file from disk to computer without playing it), they could introduce the corridor/hall reverberating echo effect by accidentally recording the output of the audio editor whilst playing and recording the disk …

Note: Do not enable “software playthrough” when recording computer playback, because this creates a series of echoes.

[* playing the audio on the disk using a media player on the computer then recording what you can hear on the computer speakers/headphones using an audio editor like Audacity]

I received the audio files again from the source after he downloaded them from his digital recorder to his computer, through It still sounds just as bad. I am going to go pick up a disk now and see if that is any clearer. Do I have any options to fix if it’s just as bad on the disk?

The person with digital recorder will be tempted to edit the file so they only send you the relevant part recording using an audio editor like Audacity. Ask them not to use an audio editor, ask them to send a copy of the whole file on the recorder which contains the relevant conversation, i.e. Connect the digital recorder to the computer and upload the audio file directly from the recorder to the

If they use some sort of audio editor (like Audacity) in the transfer of the data from digital recorder to their computer to your computer, the same reverberating echo and compression artifacts could be added again.

If the person gives you an exact copy of the file on the recorder that’s the best quality you are going to get.

An exact unedited copy of the file on the recorder may be very large: hours / 100s of Mb, with the relevant portion only a few minutes / 10s of Mb, but it’s worth the excessive download time to get a comprehensible recording. (It’s still quicker than it would take to teach your source how to use an audio editor correctly).

The whistley R2-D2 compression artifacts are incurable. There are de-reverb tools which reduce reverberation, but don’t bother as they aren’t going to help in a severe cases like yours, (and they can be horrifically expensive).