Software to separate noise from the human voice or song

Page 2 of 3 FirstFirst 123 LastLast

  1. Posts : 86
    win 10 x64
    Thread Starter

    @Bree Thanks for your help up to now. :) I did something to the file finally as follows:

    1- Using FFmpeg extracted the audio file (.mp3) of the video file (.avi)
    2- Ran the app and opened the .mp3 file
    3- Selected an amount of 8 seconds on the wave and followed the other steps of the page you'd previously offered.
    4- But in the Noise Reduction window, Step 2, I set the parameters as in the image below:

    Software to separate noise from the human voice or song-1.png

    5- Selected the entire track by "ctrl + A"
    6- Once again Effect -> Noise Reduction (There's no Noise Removal as is written in the link above!)
    7- Pressed OK. And exported the audio file as a new .mp3 file.

    Have I gone through the appropriate process correctly up to here, please?

    Then I compared that new file to the first one to see the real effect. The selected noise has properly been reduced fortunately on the whole track, but there is some annoying case, and it is, the real voices of the individuals in the track have also been rather influenced improperly, making it less comprehensible. I think this unwanted side-effect can be solved by using the parameters of the Noise Reduction window right.

    Since you've experience with the task, will you tell me what the best solution is for that, please?
      My Computer

  2. Posts : 25,032
    10 Home x64 (21H2) (10 Pro on 2nd pc)

    Well done!

    Yes, it can be a bit of an experiment to get the best parameters (and the best noise profile) I'm afraid.

    What I can suggest is that, once you have selected you noise profile, you don't have to apply it to the whole of the track. You can instead select any section of the track and apply it to just that selected part. Perhaps you could apply it to just the worst affected sections, or apply it at a low reduction level to the whole, then re-apply it to selected sections that need further noise reduction.

    You may also find some of the other effects filter useful for the sections with noticeable side effects, such as the Equalisation.
      My Computers

  3. Posts : 86
    win 10 x64
    Thread Starter

    Thank you for your guide.

    I will certainly check the Equalisation as well, but how are the parameters considered to be "a low reduction level" in your point of view, please? Something like this for example?

    Noise Reduction (dB): 10
    sensitivity: 5
    Frequency smoothing (bands): 2

    I firstly want to test it until I gradually get used to the app and the way it can best affect the audio file.
      My Computer

  4. Posts : 2,487
    Windows 10 Home, 64-bit

    I have used Audacity noise reduction on thousands of mp3s over the last 15 years or so.

    My normal procedure:

    After loading the file into Audacity, look at the waveform (or listen) and try to find a portion of the file that does not contain anything you want to keep. Only "unwanted" noise. Not a voice, not singing, not a guitar, etc. Pure noise. Often just a split second gap between words.

    That small portion can be as short as 1/10th of a second or so. If you have selected too small a portion, Audacity will put up a warning/error box, telling you to choose a longer sample.

    In an mp3 recording from a 45 rpm record or from an LP, that portion might be at the beginning or end of the song.

    You will quickly learn by looking at the wave form where that "silence" might be.

    Then open the "noise reduction" window in Audacity

    Highlight that small portion with the mouse.

    Choose "get noise profile". That will happen instantly.

    Then highlight the ENTIRE file.

    Then go back to the noise reduction menu and choose "OK".

    You will see a progress bar creep across the screen as the noise is reduced.

    I normally use a setting of 10 in the "noise reduction" box.

    I have found that the other 2 settings have little or no effect. (sensitivity and frequency smoothing).

    You can experiment by moving the noise reduction setting up to 20 or 30. You will find that those higher settings will tend to create undesirable artifacts in the exported file. I rarely use a setting above 12 or 14, and sometimes go as low as 6 or 8.

    Practice with various settings to your own satisfaction.
      My Computer

  5. Posts : 11,203
    Windows / Linux : Arch Linux

    Hi folks
    Audacity does the job but like all these things editing Audio / Video is a pain.

    I've a more engineering solution though --- do it via HARDWARE : use a recording pre-amp with decent filters built in and then save the output again -- you'll have to do this in real time but you should be able to get the sound you want -- you can also use optical for both input and output if your pre-amp has decent DAC's in it (Digital--Analog Converter's for those who aren't familiar with a lot of Audio -- BTW your mobile phone will have these in if you ever listen to music etc on a phone !!!).

    Apart from that - but just my opinion -- a lot of Modern songs -- I think I would prefer the Noise to the actual Vocals !!!!!!!

    (Actually that's just given me another idea -- I think if you play the music on a mobile phone and use say VLC - at least the Android app has really good equaliser settings - you could record on the phone !!).

      My Computer

  6. Posts : 86
    win 10 x64
    Thread Starter

    Thank you all very much.
    As for Hardware, I lack those equipment and have been somewhat used to Audacity which is well-known for the job so I had better carry on with that I think. :)

    The audio file is about 10 min and the video is a recording file and throughout all the voice you hear background noise twisted with the individuals' talks.
    I barely could separate a time amount of nearly 1.5 seconds and applied Noise 9 and 6:00, 3 for the rest. The outcome is good and lots of noise is reduced.

    Now how can I enhance the quality of the voice in it? Is it done by Equalisation? If so, how please?

    I don't know what to do with this window!

    Software to separate noise from the human voice or song-3.png

    Whatever is to be done should be done for the whole file.
      My Computer

  7. Posts : 245
    W10 Home Version 21H1 Build 19043.1055

    Voice EQ needs a slight lift (+) around 2-5 Khz, if it sounds a bit boomy then roll off the bass end under 100 Hz and/or see if there is a 'high pass' filter which will do a similar thing.
    If it sounds 'boxy' then you need to cut (-) around the 250-400 Hz mark.
    You should only need to do around 1-3 db of boost/cut unless there are major issues in which case more targeted 'surgical' EQ (parametric with variable Q) would be better.
    Last edited by Infrasonic; 15 Sep 2019 at 07:31.
      My Computers

  8. Posts : 2,487
    Windows 10 Home, 64-bit

    tomyfr said:

    Now how can I enhance the quality of the voice in it?
    Equalization and filters are about all you have if the "noise" is largely gone.

    There's a lot of trial and error involved---experimentation.

    Don't get your hopes too high. You are severely limited by the quality of your original file. Hard to overcome that.

    Is this a stereo file---different sound in each speaker?

    If so, you might have some luck by splitting the stereo track into 2 separate tracks and then using Audacity on the best of those 2 tracks. This can work if the left and right channels are significantly different.

    Or you might try to export the stereo track, saving it as a mono track...........and then reopening that mono track in Audacity and operating on it.

    Lots of possibilities, but you have to be willing to fiddle and see what happens.

    Is this an mp3 or a WAV or what?

    What type of "noise" did the original file have?? Hiss? If so, the Audacity noise remover is about as good as it gets.

    Isolated ticks or constant crackling like frying bacon? If so, it might be worthwhile to export your original as a WAV file and then use something other than Audacity to attack that noise. Some of the better "click and pop" removal tools will operate ONLY on a WAV file. At times, I've converted an MP3 into a WAV and then used a click tool to great effect, and then converted it back to an MP3.
      My Computer

  9. Posts : 86
    win 10 x64
    Thread Starter

    The issue with the voice is that it has so much distortion making it hard to understand.
    And unfortunately I couldn't comprehend the instructions above for using Equalization! Does it do something good for that problem of voice?

    - - - Updated - - -

    Let me please share the audio file with you just after being separated from the video by the audacity app (without any effect being done yet)

    Here it is.

    Now how to best remove the noise from it and enhance the quality too?
    Please write the steps you apply for the job, here, so that I can learn and test it for other audio files of those video files too. The language, however, is not English.
      My Computer

  10. Posts : 245
    W10 Home Version 21H1 Build 19043.1055

    tomyfr said:
    The issue with the voice is that it has so much distortion making it hard to understand.
    And unfortunately I couldn't comprehend the instructions above for using Equalization! Does it do something good for that problem of voice?
    No if the issue is distortion then EQ won't help really. Is the level of distortion present in the original file or is it getting worse after you play around with it?
    If it's getting worse then there may be gain clipping in the software chain (the parameters you are using may be too extreme).
    It's difficult to answer that remotely because obviously I can't hear what's going on.
      My Computers


  Related Discussions
Our Sites
Site Links
About Us
Windows 10 Forums is an independent web site and has not been authorized, sponsored, or otherwise approved by Microsoft Corporation. "Windows 10" and related materials are trademarks of Microsoft Corp.

Designer Media Ltd
All times are GMT -5. The time now is 15:51.
Find Us

Windows 10 Forums