Hi all,

Something went wacky during the recording and I ended up with noise in my recording. I removed the noise with Izotope RX4 but I think the audio needs more improvement. Can anyone listen to these 3 recordings and help me figure out what other steps I can or should take to improve it?

The first recording is the original, the second is one level of noise removal and the third is two levels of noise removal. Thanks a lot.
https://www.dropbox.com/sh/gotzwvrxuot84wl/AAATGGUgVDRdiXG-H6d2h_MZa?dl=0

Kathy

Regarding the original, I don't think noise is a huge problem (around -43dBFS), the tonal quality is.. sounds telephone like and sibilant with no body. Possible broken or wet mic, impedance issue and/or improperly wired connector(s)?
BTW, A mono file would suffice in this case and would be half the size and download duration... If this is for video, why is the file 44.1kHz. (not that it would make any difference in the quality). I'll check the other files later on when I have more time to DL.

Regarding the original, I don't think noise is a huge problem (around -43dBFS), the tonal quality is.. sounds telephone like and sibilant with no body. Possible broken or wet mic, impedance issue and/or improperly wired connector(s)?
BTW, A mono file would suffice in this case and would be half the size and download duration... If this is for video, why is the file 44.1kHz. (not that it would make any difference in the quality). I'll check the other files later on when I have more time to DL.

Thanks Rick. I extracted the audio from the video without paying much attention to what the export settings were so probably that's why you are seeing 44.1kHz
The other two files have noise removed but I did nothing to improve the quality of the voice.
Thanks for looking into it.

Broad-band "white noise" is the very worst kind of noise to attempt to remove. Because it completely overlaps the "band of interest", the speech recording. However, Izotope RX4 did an amazing job of removing it.

Since we know nothing about the equipment or setup you used, it isn't really possible to attempt to analyze the CAUSE. The poor signal-to-noise ratio may have been from recording at a level that was too low. The tonal quality of the speech is a different matter. It could have been caused by a dozen different things including poor technique poor mic selection, and even defective equipment.

Since Izotope did such a remarkable job of removing the hiss, perhaps simply applying some judicious EQ would improve the tonal quality. I would think that a parametric "dip" at the offending frequency would do wonders for removing that "tinny" or "nasaly" quality.

Broad-band "white noise" is the very worst kind of noise to attempt to remove. Because it completely overlaps the "band of interest", the speech recording. However, Izotope RX4 did an amazing job of removing it.

Since we know nothing about the equipment or setup you used, it isn't really possible to attempt to analyze the CAUSE. The poor signal-to-noise ratio may have been from recording at a level that was too low. The tonal quality of the speech is a different matter. It could have been caused by a dozen different things including poor technique poor mic selection, and even defective equipment.

Since Izotope did such a remarkable job of removing the hiss, perhaps simply applying some judicious EQ would improve the tonal quality. I would think that a parametric "dip" at the offending frequency would do wonders for removing that "tinny" or "nasaly" quality.
Thanks Richard. I don't really care to find out what happened. This is not the equipment I normally use and everything was a last minute, I'm sure given time I would have figure it out on the spot.
Do you think I should try to improve the first processed file or the second one?

The final (double-processed) recording sounded good enough (except for the "tonal balance")
I would simply play with EQ on your final track.

It is true that in some cases certain types of processing should be done in a particular sequence.
But I don't believe that is the case in this particular situation because the result from the Izotope processing is so good.

Thanks everyone. I tried to improve this audio but I'm not sure I succeeded. Can someone give it a listen and let me know what they think?
Thanks a lot. I really appreciate all the help here.

https://www.dropbox.com/s/qkjcp9zybjc4o4n/LaraProcessed4.wav?dl=0

Better. It still sounds overly sibilant to my ear, but that could be an artifact of my listening environment (Sony MDR-7506 headphones in my office cubicle). Fixing that would be simply a matter of rolling off the excessive high-frequency response.

Did you try "sweeping" a notch filter through the speech frequency band? (300 Hz ~ 3KHz) It might be very instructive even if you don't end up using it.

I don't remember whether you mentioned what editing software you are using (or have access to). I use Adobe Audition (formerly Cool Edit Pro) and it has a nice parametric EQ feature that lets you experiment with the peak/notch in real-time. You can control the position (frequqncy), the height/depth (boost or cut dB), and the sharpness ("Q") of the peak or notch.

I would try for a bit more of a notch around the resonant frequency band. I will try it when I get back to my edit system after our concert this evening.

PS: Please note that you don't need to provide such a loooooooooooooooong sample!!! 30 seconds would be more than adequate in most cases. Much better a short and uncompressed sample than a long, compressed sample.

Better. It still sounds overly sibilant to my ear, but that could be an artifact of my listening environment (Sony MDR-7506 headphones in my office cubicle). Fixing that would be simply a matter of rolling off the excessive high-frequency response.

Did you try "sweeping" a notch filter through the speech frequency band? (300 Hz ~ 3KHz) It might be very instructive even if you don't end up using it.

I don't remember whether you mentioned what editing software you are using (or have access to). I use Adobe Audition (formerly Cool Edit Pro) and it has a nice parametric EQ feature that lets you experiment with the peak/notch in real-time. You can control the position (frequqncy), the height/depth (boost or cut dB), and the sharpness ("Q") of the peak or notch.

I would try for a bit more of a notch around the resonant frequency band. I will try it when I get back to my edit system after our concert this evening.

PS: Please note that you don't need to provide such a loooooooooooooooong sample!!! 30 seconds would be more than adequate in most cases. Much better a short and uncompressed sample than a long, compressed sample.

Thanks Richard. I have Adobe Audition as well as Izotope Ozone 5. But I have to admit I don't know how to really use them. I'm not really sure how to apply "sweeping" a notch filter or how to use parametric EQ. If you can let me know how to do it that would be great.

Better. It still sounds overly sibilant to my ear, but that could be an artifact of my listening environment (Sony MDR-7506 headphones in my office cubicle). Fixing that would be simply a matter of rolling off the excessive high-frequency response.

Did you try "sweeping" a notch filter through the speech frequency band? (300 Hz ~ 3KHz) It might be very instructive even if you don't end up using it.

I don't remember whether you mentioned what editing software you are using (or have access to). I use Adobe Audition (formerly Cool Edit Pro) and it has a nice parametric EQ feature that lets you experiment with the peak/notch in real-time. You can control the position (frequqncy), the height/depth (boost or cut dB), and the sharpness ("Q") of the peak or notch.

I would try for a bit more of a notch around the resonant frequency band. I will try it when I get back to my edit system after our concert this evening.

PS: Please note that you don't need to provide such a loooooooooooooooong sample!!! 30 seconds would be more than adequate in most cases. Much better a short and uncompressed sample than a long, compressed sample.
Hi Richard

I wonder if you had a chance to play with my file. I don't seem to be able to make it any better. I'm getting kind of desperate here, haha.

Kathy

Hi, Kathy.
I experimented a bit with the parametric EQ in Adobe Audition, and this is my first-pass attempt at equalizing the dialog.

First, I selected a part of the timeline that had something approaching a complete paragraph(!) I picked the section from 1:00:01 to 1:00:46. You can't EQ the on-mic subject and the off-mic interviewer together, IMHO.

Then I clicked "Effects" in the menu, and then selected "Filter and EQ", and picked "Parametric Equalizer", to bring up the window shown here.

First, I rolled off the high-end which seemed quite excessive even to my old, tired ears. I dragged down the high-frequency control to -22.4 dB as shown.

Then I turned on the #4 filter and dragged it to around 3590 Hz, and down to -13.3 dB to try to tame what I heard as an annoying resonance.

Of course, you use the "play button" in the lower left corner so you can hear the track while you are playing with the EQ. And you can turn the EQ on and off using the green button next to the "play button".

Now this is only what I heard with my ears and my JBL LSR305 speakers. Feel free to turn on other frequency bands and drag the nodes around to "season to taste" for yourself.

I sometimes drag the EQ points UP above the center line to emphasize the frequency just to try to make it "worse" so that I can "tune" it to the offending frequency/band. Then I drag it DOWN to filter it out of the track to whatever degree seems appropriate.

Certainly this is all VERY SUBJECTIVE. I tried to demonstrate HOW I did it rather than declaring "this is how it should be done" because that is ultimately your decision (and your client's).

Hi, Kathy.
I experimented a bit with the parametric EQ in Adobe Audition, and this is my first-pass attempt at equalizing the dialog.

First, I selected a part of the timeline that had something approaching a complete paragraph(!) I picked the section from 1:00:01 to 1:00:46. You can't EQ the on-mic subject and the off-mic interviewer together, IMHO.

Then I clicked "Effects" in the menu, and then selected "Filter and EQ", and picked "Parametric Equalizer", to bring up the window shown here.

First, I rolled off the high-end which seemed quite excessive even to my old, tired ears. I dragged down the high-frequency control to -22.4 dB as shown.

Then I turned on the #4 filter and dragged it to around 3590 Hz, and down to -13.3 dB to try to tame what I heard as an annoying resonance.

Of course, you use the "play button" in the lower left corner so you can hear the track while you are playing with the EQ. And you can turn the EQ on and off using the green button next to the "play button".

Now this is only what I heard with my ears and my JBL LSR305 speakers. Feel free to turn on other frequency bands and drag the nodes around to "season to taste" for yourself.

I sometimes drag the EQ points UP above the center line to emphasize the frequency just to try to make it "worse" so that I can "tune" it to the offending frequency/band. Then I drag it DOWN to filter it out of the track to whatever degree seems appropriate.

Certainly this is all VERY SUBJECTIVE. I tried to demonstrate HOW I did it rather than declaring "this is how it should be done" because that is ultimately your decision (and your client's).
Thanks Richard. I will listen to it when I get home. I forgot to mention that I do not care for the interviewer's voice. It will not be used.

Hi, Kathy.
I experimented a bit with the parametric EQ in Adobe Audition, and this is my first-pass attempt at equalizing the dialog.

First, I selected a part of the timeline that had something approaching a complete paragraph(!) I picked the section from 1:00:01 to 1:00:46. You can't EQ the on-mic subject and the off-mic interviewer together, IMHO.

Then I clicked "Effects" in the menu, and then selected "Filter and EQ", and picked "Parametric Equalizer", to bring up the window shown here.

First, I rolled off the high-end which seemed quite excessive even to my old, tired ears. I dragged down the high-frequency control to -22.4 dB as shown.

Then I turned on the #4 filter and dragged it to around 3590 Hz, and down to -13.3 dB to try to tame what I heard as an annoying resonance.

Of course, you use the "play button" in the lower left corner so you can hear the track while you are playing with the EQ. And you can turn the EQ on and off using the green button next to the "play button".

Now this is only what I heard with my ears and my JBL LSR305 speakers. Feel free to turn on other frequency bands and drag the nodes around to "season to taste" for yourself.

I sometimes drag the EQ points UP above the center line to emphasize the frequency just to try to make it "worse" so that I can "tune" it to the offending frequency/band. Then I drag it DOWN to filter it out of the track to whatever degree seems appropriate.

Certainly this is all VERY SUBJECTIVE. I tried to demonstrate HOW I did it rather than declaring "this is how it should be done" because that is ultimately your decision (and your client's).
Hi Richard
My Parametric Equalizer in Adobe Audition doesn't look like that. Is this some kind of plugin?

It is the one that came with Audition. I have not installed any extra plug-ins.
What does yours look like? Doesn't it have the same basic functions?
The window may not look identical, but the basic function has been the same through almost all of the versions.

It is the one that came with Audition. I have not installed any extra plug-ins.
What does yours look like? Doesn't it have the same basic functions?
The window may not look identical, but the basic function has been the same through almost all of the versions.

Hi Richard,

My Parametric EQ looks like this and it doesn't say VST plugin like yours does.

I am using the "Creative Cloud" subscription version of the Adobe applications.
I don't know which version of Audition you are using, but the basic functionality is the same.
Simply drag that "4" spot over to somewhere around 3600 Hz and down ~12dB
And the "H" spot over on the right end controls the low-pass (high-cut) filter as I described.
You can simple mouse-over and drag those filter spots around and hear what they sound like.

VST is a standard for using plug-ins from different vendors.
Apparently Adobe uses the VST standard even for its own built-in filters..
For the purposes of this discussion, it has no significance.

I am using the "Creative Cloud" subscription version of the Adobe applications.
I don't know which version of Audition you are using, but the basic functionality is the same.
Simply drag that "4" spot over to somewhere around 3600 Hz and down ~12dB
And the "H" spot over on the right end controls the low-pass (high-cut) filter as I described.
You can simple mouse-over and drag those filter spots around and hear what they sound like.

VST is a standard for using plug-ins from different vendors.
Apparently Adobe uses the VST standard even for its own built-in filters..
For the purposes of this discussion, it has no significance.

Hi Richard,

I also use the CC version of Adobe Audition and don't have the same parametric EQ but it doesn't matter.
I did what you suggested and it sounds very muffled to me. I listened on two different sets of headphones (Sennheiser HD 280 Pro and Sennheiser HD 800).
I think I need to add some frequencies back in now.
https://www.dropbox.com/s/afzzbx3z1ty5bk5/LaraProcessed4_01.wav?dl=0

I did what you suggested and it sounds very muffled to me. I listened on two different sets of headphones (Sennheiser HD 280 Pro and Sennheiser HD 800).
I think I need to add some frequencies back in now.
I was not RECOMMENDING the settings I mentioned. I was only demonstrating HOW to do it. Please remember that what YOU hear with YOUR monitoring system, and what YOU think sounds "OK" may be very different. So I STRONGLY encourage you to experiment for yourself. DO NOT use the same settings that I used! Work out your own. That is what I meant by "season to taste".

It is poor technique to remove something that you will subsequently restore in a subsequent step. It is guaranteed to inflict audible trauma to your track.


https://www.dropbox.com/s/afzzbx3z1ty5bk5/LaraProcessed4_01.wav?dl=0
PLEASE do not post hour-long samples!!! 30-60 seconds of an appropriate section will be MORE than enough for anyone to hear what you are talking about.

I've got RX4 Adbanced with a set of Event 2020 monitoring speakers.

I've done a couple of passes of broadband noise reduction and this in turn shows up the 'artificial fuzz' that I'm guessing is possibly the result of a low quality process that has been applied to your original file at some stage, possibly in the way you got it out of the recording gear.

I've tried some "dereverb" (aka echo removal) and a stiff application of this has really helped. There is still problems with the audio that remains. It's just not going to be the perfectly usable recording that you would want.

If you have the advanced version of RX4, you can grab the text file that I have uploaded, rename to a .xml extension type, and then import this in to Izotope RX4 and see what I mean.

To be honest, I'm not sure how you are going to be able to satisfactorily fix this one. You might want to have another crack at getting the audio off the recording equipment exactly as it was recorded in case this makes a difference.

Andrew

Do you have a good recording of this same person?

I was not RECOMMENDING the settings I mentioned. I was only demonstrating HOW to do it. Please remember that what YOU hear with YOUR monitoring system, and what YOU think sounds "OK" may be very different. So I STRONGLY encourage you to experiment for yourself. DO NOT use the same settings that I used! Work out your own. That is what I meant by "season to taste".

It is poor technique to remove something that you will subsequently restore in a subsequent step. It is guaranteed to inflict audible trauma to your track.


PLEASE do not post hour-long samples!!! 30-60 seconds of an appropriate section will be MORE than enough for anyone to hear what you are talking about.
Ooops sorry, I thought these were the settings you used and got good results.

I've got RX4 Adbanced with a set of Event 2020 monitoring speakers.

I've done a couple of passes of broadband noise reduction and this in turn shows up the 'artificial fuzz' that I'm guessing is possibly the result of a low quality process that has been applied to your original file at some stage, possibly in the way you got it out of the recording gear.

I've tried some "dereverb" (aka echo removal) and a stiff application of this has really helped. There is still problems with the audio that remains. It's just not going to be the perfectly usable recording that you would want.

If you have the advanced version of RX4, you can grab the text file that I have uploaded, rename to a .xml extension type, and then import this in to Izotope RX4 and see what I mean.

To be honest, I'm not sure how you are going to be able to satisfactorily fix this one. You might want to have another crack at getting the audio off the recording equipment exactly as it was recorded in case this makes a difference.

Andrew

Unfortunately, I do not have the advanced version of RX. This is how the track sounds. It has nothing to do how I got it off the recording equipment. I know it won't sound great but I want to make the set out of it.

Do you have a good recording of this same person?
Hi Gary,

No, I don't, unfortunately. I could get it if it's helpful.

I could get it if it's helpful.

I for one would be making a beeline for getting the good copy of the recording.

Why are we otherwise doing all this messing around here???

Andrew

I for one would be making a beeline for getting the good copy of the recording. Why are we otherwise doing all this messing around here???

Andrew, did you not get a chance to hear the original samples? The original recording is "down in the mud" with a bad noise-floor. Two passes with iZotope did a miraculous job of getting rid of most of the noise. However the very poor tonal balance is a different matter.

There is no "good copy of the recording". Unless you are implying that they re-record the whole thing?

The poor tonal balance could be significantly improved by judicious use of a parametric equalizer and I was trying to show how to do it. However it does seem pretty resistant to getting decent balance out of it.

I thought she was indicating that there was a better quality of the audio that she could go and get. We might need to wait for clarification from her on that one.

Andrew

I thought she was indicating that there was a better quality of the audio that she could go and get. We might need to wait for clarification from her on that one.

I thought Gary asked if I had any voice recording of this person. I thought he wanted to compare it or something like that. I don't have a second recording of the interview and I can't redo it

Ahh. I see.

All the best with making the most of what you have.

Andrew

I thought Gary asked if I had any voice recording of this person. {snip} I don't have a second recording of the interview and I can't redo it

I (Greg, not Gary) understand that this is the only recording of this interview. I want to know whether you have any other good recording of this same person's voice.

I (Greg, not Gary) understand that this is the only recording of this interview. I want to know whether you have any other good recording of this same person's voice.
Sorry GREG. I do not have any other recording of the person but I can get it. Should it be from the same location?

OK, what do you all think of this version? (I shortened it so you don't have to download this long file):

https://www.dropbox.com/s/j74z81ste24ysht/Lara%20_Processed_Short.wav?dl=0

Thanks

That's better than I was expecting to hear in terms of the tonality of her voice. There is a little background noise there but if you had just a tiny bit of music (or actuality) underneath her whilst she is speaking, I reckon you might just get away with it as it is.

All is not lost.

Andrew

That's better than I was expecting to hear in terms of the tonality of her voice. There is a little background noise there but if you had just a tiny bit of music (or actuality) underneath her whilst she is speaking, I reckon you might just get away with it as it is.

All is not lost.

Andrew

Thank you. Yes, there will be music.

Sorry GREG. I do not have any other recording of the person but I can get it. Should it be from the same location?

Another recording of the same person, ideally in the same location, might help us come up with appropriate EQ to make her sound "right." Of course that assumes that everyone involved feels that "reference" recording sounds acceptable.

As it is now, we're just guessing at how her voice ought to sound, and any EQ we apply (based on that guess) might end up making the background noise worse. Personally, I would like to understand the EQ situation first, before I decide how to approach the NR process.

This is going to be a very tough track, because sometimes the talent is really projecting her voice to the group, and other times she trails off and is almost mumbling to herself. I don't know if you will ever get the track clean enough so that the mumbled parts are useable.