Audio file anomalies

[mr_es335]

Good day,

I have a number of audio files that were converted by the authors from tape [?] to .mp3 format.

These audio files consist of 1) very soft spoken voice, and 2) various music inserts.

The main "issue" that I am presently experiencing is the with regards to the contrast in volume between the voice and the music - with the music portion, obviously, being considerably louder than the spoken word.

What can be done - if anything, to alleviate this anomaly?

• PS #1: I do have access to the Levelizer and the Studio Frequency Analyzer - if this is of any assistance.
  
• PS #2: I am also willing to pay to have someone explain to me how to accomplish this task - as there are approximately 16 data files that require such processing - with more coming.
  
• PS #3: Interested parties may "remote-in" to my system via "AnyDesk" as well - if this will be helpful.

Thanks to any that respond...very much appreciated.

[John Ludlow]

Does the voice exist within a fairly consistently narrow frequency band? Can you bring up its comparative volume by identifying that frequency slice and raising its volume relative to the rest using the parametric equalizer?

[mr_es335]

John,
Thanks for the response...much appreciated...

This entire process is rather foreign to me.

• For example, I hire someone here to configure my live rig - configuring EQ and such...as all that I prefer to do is to play.
• I only perform in three venues, so I have presets for each venue which works out just fine for me.

So, I guess I looking for a bandaid-type approach.

[John Ludlow]

John,
Thanks for the response...much appreciated...

This entire process is rather foreign to me.

• For example, I hire someone here to configure my live rig - configuring EQ and such...as all that I prefer to do is to play.
• I only perform in three venues, so I have presets for each venue which works out just fine for me.

So, I guess I looking for a bandaid-type approach.

I'm not sure I understand, as the above doesn't seem to jibe with your original request (it's probably me...). I had understood that you have several recordings that have both music and voice on them - and you want to raise the level of the voice relative to the music. If this is the case, and the frequency range of the voice is more limited, or other than, the frequency range of the music, then one of the parametric equalizers built into SAW might be perfect to solve the problem.

First you would convert the files from .mp3 to .wav files using one of the hundreds of music format converters available online.

That will allow you to bring them into SAW one at a time, and play them back.

You probably already do, but just in case you don't know what a parametric EQ is, or how it works: it is a frequency filter, just like a set of tone controls on a stereo. On a stereo, maybe you have 'Hi', 'Mid', and 'Low'. Each of those represents a subset of the entire frequency range. When you turn one up higher or lower than the others, that frequency range is made louder or softer. But, of course, its usefulness is limited to situations in which adjusting a third of the available range is what is needed. Similarly, graphic equalizers, which typically have a bunch of sliders corresponding to different frequency ranges (usually octaves, or portions of octaves...) are identical in function to low/mid/hi - except the entire frequency range is divided into more sub-ranges. But, again, an octave might be too large, or small, an amount for your application. And this is where parametric equalizers come in.

A parametric EQ is not limited to a pre-set frequency range width. So, you first define the center frequency that you want to affect, then you define the 'width' of frequencies, surrounding that center frequency that you want to affect, and then you turn its volume up or down.

In this application, you'll want to roughly identify the center frequency of the voice by experimentation. It will be somewhere in the midrange area. Start by setting the frequency range width to be, say, half an octave (octaves are exactly double the frequency from each other; so an octave from 'A' 440 is 'A' 880, and the next is 'A' 1760). Turn, the volume for that defined range up. Then, while you play back the recording, and while leaving the width (also called the 'Q') alone, change the midpoint of the center frequency until the voice is more clear. As you continue to adjust the midpoint past that, it should become less clear - that lets you know that you have passed by the frequency range of the voice. Then, you'll want to adjust backwards until the voice is louder again.

From there, you'll want to narrow the 'Q' a little, and move the center frequency higher or lower until the voice is most clear again (indicating that it has entered into the range controlled by your parametric EQ settings). Do this several times until when you narrow the Q part of the range of the voice is removed. Then open the Q back up again until the voice is neatly within its range. Finally, turn the volume up on the parametric until the balance of music to voice is what you prefer.

And then, output that to a new wave file and the new file will contain what you have modified.

The caveat here, Dell, is that the music might also be using the same frequency range that the voice is. In that case, whenever you raise the volume for the voice, the volume of the music will also go up - thus continuing to obscure it. But, it's worth a try and should only take a few minutes.

Also, if that won't work then I don't know what will (but I am a junior sound engineer at very best).

[MMP]

Izotope has a module in RX9 that can create stems from a stereo or mono file and split the vocal out separately from the musical elements. I’m not sure how well it might work with spoken material. I could try it out for you if you care to send a file. Mmpro1 at mac dot com.

[mr_es335]

Good day,

Thanks for the replies...again very much appreciated!!

The narrations are - I would say very consistent...it is just the music portions that vary.

What I am suggesting is the following:

1. I send 2-3 audio flies
2. Configure the Levelizer to focus on the narration and "amplify" that portion - if this is indeed what is required?
3. Save the configuration as a Levelizer preset
4. Send me that Levelizer preset
5. Tell how much you would charge for this service?

For your interest, I have contacted the provider to make them aware of this anomaly.

That is pretty much it!

Thanks again!!

[radiorog]

Hi, Dell.

I'm a little unclear about whether the narration and music are individual segments of the audio, or whether they're mixed (music underlaying the voice).

You say the music portions are 'inserts', which suggests that they don't overlay the narration. If that's actually true, you could just cut the narration out of the original track, move those regions to a separate track, and increase the volume (or decrease the music volume). It's a bit of work, but seems simple enough.

Of course, if the narration and music are together, you'd need to try one of the other suggestions.

Just a thought ...

Roger

[mr_es335]

Good day,

Here is a link to one of the audio files: I.O.U.

Notes
It would appear that at around 2:10 the music actually decreases a bit - probably manually, and continues at this new reduced level till the end of the song.

• I guess it would be possible to "ride a fader" during the louder music parts - as was done here.
• I was hoping for a more automatic route however - if this is at all possible?
• Music quality is not that important - it is the narration that I am desiring.

And John, if you are, as you say, "...a junior sound engineer at very best..." ..then I am still in diapers - when in comes to such things.

Thanks again for those that have offered their assistance!

[John Ludlow]

If you haven't yet, I think you should send some files to Michael, a real sound engineer, if he's willing to help. It sounds as if, being a professional, he has a program that is made to do just what you want done. I still think my idea might work. But for a solution where you don't have to do or learn anything - it sounds to me like Michael's your guy.

[mr_es335]

John,

I have provided a link to the data file above.

Please view: Post #6