Topic: Restoring/enhancing audio recording

[miwilker]

Hello all, I'm looking for some help/advice and a friend told me you all might be a good resource...

I just completed a phone interview for a book I’m working on and discovered that the audio of the interviewee’s voice is barely audible.  I can’t make out what she’s saying, even with the volume turned all the way up (which is strange, because the actual audio in my earpiece during the conversation was fine).
 
I have some confidence that the audio is recoverable because I can hear her voice.  It’s just at such a low level that it’s unintelligible.  I downloaded the free ‘Audacity’ program and amplified her voice, but there’s a substantial hiss which also was amplified so her speech remains unintelligible.  I used the Audacity 'noise removal' tool but it wasn't effective (it degraded her voice along with removing some of the hiss).  I need to find out if it’s possible for someone to amplify her voice and remove that hiss without degrading her voice.  All I need is to be able to understand what she’s saying, so I can transcribe it.
 
I’m wondering what options I have as to getting the sound back to an audible level.  It’s an 8MB WMA file… about an hour long conversation.  Then there’s another one which is about 10 mins long… 1.4MB.
 
Any ideas / advice would be greatly appreciated.

Thanks

Mark Wilkerson

[voltronic]

I've never had much success with Audacity's noise removal either.  You need something that's purpose-built for this task (but definitely not free though):

http://www.izotope.com/products/audio/RX/

Or, if you are willing to post a link to the files here maybe one of us that has RX can work on this for you...

[page]

I've never had much success with Audacity's noise removal either.  You need something that's purpose-built for this task (but definitely not free though):

http://www.izotope.com/products/audio/RX/

This.

RX is the best tool I can think of for this sort of task.

[flipp]

a free tool you might try: levelator though you may need to convert the file to a different format for it to work

from their page

What is The Levelator®?

Do you believe in magic? You will after using The Levelator® to enhance your podcast. And you'll be amazed that it's free, now even for commercial use.

So what is The Levelator®? It's software that runs on Windows, OS X (universal binary), or Linux (Ubuntu) that adjusts the audio levels within your podcast or other audio file for variations from one speaker to the next, for example. It's not a compressor, normalizer or limiter although it contains all three. It's much more than those tools, and it's much simpler to use. The UI is dirt-simple: Drag-and-drop any WAV or AIFF file onto The Leveler's application window, and a few moments later you'll find a new version which just sounds better.

Have you ever recorded an interview in which you and your guest ended up at different volumes? How about a panel discussion where some people were close to microphones and others were not? These are the problems the post-production engineers of Team ITC here at The Conversations Network solve every day, and it used to take them hours of painstaking work with expensive and complex tools like SoundTrack Pro, Audacity, Sound Forge or Audition to solve them. Now it takes mere seconds. Seriously. The Levelator® is unlike any other audio tool you've ever seen, heard or used. It's magic. And it's free.

< Appears the download links may no longer be available. If you need the windows or Linux versions, let me know as I have the most recent of both saved on a different comp and can upload one/both to one of my webpages. >

[miwilker]

Thanks for the replies all - great info.

Mark

[TSNéa]

Hello miwilker,

I understand your goal is only to get something you can hear clearly enough to listen to and understand the interviewee. We're not dealing with music!

I'm not sure about Levelator, I think I used it before but it acted more like a compressor, maybe a smart compressor but a compressor. Try it!
And I heard how effective RX is but I can't afford it. Maybe when I make my living with audio?

The point is to understand "where" is your hiss... If it is low enough under the level of the voice, one (very) basic way to go is to use some kind of noisegate/expander/whatsoever: you cut everything under a threshold you carefully choose to get the best result. Or you expand everything above a threshold.
If the hiss is "mixed" with the words: RX, OK; Audacity: you have to select a part of the recording showing only the hiss, without any speech and "teach" Audacity that that's what you want to eliminate. I some occasions I used a free (I don't remember buying it) audio editor software which was able to clean the sound of a poorly recorded lecture/interview: WavePad by NCH Software.
http://www.nch.com.au/wavepad/

Open your file, select a sample to test, go to the Effects menu, then to Noise reduction, then to Automatic spectral substraction? (I'm sorry: I have a French version...). You can try with the default settings or tune the silence vs audio ratio or apply one of the presets : voice or music.
And you have three others types of noise reduction to try...
It looks like Audacity, with more options. And I think you can find this feature in other free editors.

I hope this will help you.

[miwilker]

Hi all, here's a snip of what I'm dealing with. 

https://www.dropbox.com/s/cypvupydxlf2pl7/Sample.wav

Nobody's been able to get the audio to an intelligible level yet... is this material just too poor to work with?  Further thoughts / advice would be greatly appreciated.

Thanks

Mark

[furburger]

day late to the party, but I had to fiddle with the noise removal part for all of 20m to *completely* clean up a Gary Clark Jr. show from a blues tent in Memphis..there was a 'hum' in the background, and it's now gone.

the key for me was to isolate just a fraction of a second for the 'noise profile' (more time - more noises for it to add to its filter), and make sure there was no crowd noise...surely there's a quiet part between speaking?

and if you even want to 'loophole' it, where you boost the voice before you apply the noise 'removal' (think 'reduction', as removal would be silence):

1. load OG file
2. get the 'noise profile'
3. open the same file in another tab.
4. boost it, warts and all.
5. delete the file in the 1st audacity window with the noise profile you created, but DON'T close the tab it's in (go to 'file', then 'close' on the left vs. the red X)
6. go back to 2nd opened file, copy
7. paste it into the OG audacity window, which still has the noise profile you created
8. go back to effects, and at the very top it should say something about 'noise profile' (vs. in alphabetical order), click that
9. wait
10. enjoy your results


that way, you're applying the noise removal POST boost, without having to get the profile from the boosted track. more of YOUR noise gets thru, less of the hiss.


will try it on your sample, and PM you if it works (I can hear what she's saying, it's not easy, but I'll transcribe what I heard)

[Bruce Watson]

Hello all, I'm looking for some help/advice and a friend told me you all might be a good resource...

I just completed a phone interview for a book I’m working on and discovered that the audio of the interviewee’s voice is barely audible.

I don't know if this will help with your current situation, but it might help you with future recordings of phone calls: Ty Ford talks about how to improve audio for phone / skype call interviews. Ty Ford is pretty well regarded in the audio world, but I don't think he's a Taper. If you decide to pursue this with him, Ty is a moderator of the CreativeCOW Audio Forum; if you ask there you have a good chance he'll be one of the responders.

OTOH, there's nothing quite like the level and diversity of experience to be had with the Tapers here. Somebody here probably has the answer. Sadly, it's not me.