Hi,
I have a recorded phone conversation of which I need to raise the voice that was coming out of the phone speaker (I imagine it was recorded with a handheld recording device held up close to the receiver?).

This might be really basic stuff for a lot of you but I'm very new to editing in this way and tried just putting a compressor (which I'm not sure I know how to use properly either?!) on the track, but it raise the other 'local' voice to the point of distortion!

I'd just like to have both voices at +- the same level.

Thanks in advance!

 

I would first use waves ns-1 to clean up the recording, probraly at 0.1 and a little bit of gain on the track,
if possible without clipping. (ns-1 is a real butcher.)

then I would duplicate the track in any daw, slice them up exactly at the same spots.
lower the gain on one lane till it has the desired volume.
( I would keep the seemingly "silence" parts of the wavform included in the original track and not the lowered duplicate)

next I would eq and compress, the order does not really matter unless you raise the gain on the highs ( Sibilance will make the eq go first)
with a little tweaking you can tune in the desired atmosphere by tweaking waves ns-1 (or at least the volume that it's fet)

This is actually pretty easy to set up in fl studio, I could imagine not every daw makes it that easy wich might make the ns-1 part a back an forward.

I could however make a fl studio 12.5 (build 59) projects with a little instructions (additional waves plugins required)

 

By manually editing the whole thing and adding gain only to what you want louder

 

Keep all the noise. Most noise reduction plugs have a "." keep that on a separate track & mix to taste.

 

Thanks....although, I was hoping someone would point me to some awesome plugin that does that in a few clicks!
I seemed to remember seeing an "isolation" software that sort of did what a spectral editor does with frequencies but did it with a dynamic range.
Because it's a really simple recording. You really just have 3 elements in it: A very loud voice (speaking on the phone), an almost inaudible voice (coming from the phone receiver) and the residual room noise. The noise is easy to get rid of, so it's really just 2 elements, the 2 voices.
And as I said, in my mind, I can remember seeing this video demo of this software that would isolate sound according to dynamic range. Does that ring a bell with anyone?

 

If you want to isolate just_quiet from just_loud, a simple gate would do. It will sound awful, like an old police/CB radio cutting in/out with the squelch knob. You'd be better off simply automating the volume, like @Bill Vkerchi? suggested. That's not without its own problems (the background noise will get amplified along with the parts you turn up, so would sound unnatural), but quick & painless (just ride the volume & record automation).

 

I would also use a noise-profiling denoiser, capture the profile of *the other voice* and reduce that level, which is effectively what you want.

 

If it's a one-time thing and not top secret material, feel free to send it to me. I'm thinking I'll do it with some finetuned spectral compression/expansion (Melda plugs), but the process of tuning it is by ear and would take an age to explain.

WAVs or FLAC can be sent via WeTransfer.com

 

Spectral editors like Spectralayers or Izotope Rx will help you identify visually the noise and the various voices. You can then use the dedicated tools to raise a particular voice or lower/eliminate the noise or even extract the voices from the content. It is not the easiest of tasks, if you haven't used such an editor before it takes sometime to get to grips, but imho it has the best results.

 

Awesome! That's really kind of you Satai, thanks
It's not top secret at all. It's kind of an endearing story actually...it's for my 87 year old dad. It's from an old audio K7 of my deceased mother who's having a phone argument with some worker regarding a piano transportation issue. My dad just loves to hear her voice and her arguing skills!
I've already done a few easier tapes where it really just involved removing room noise and tape hiss, but this is a lot trickier, for me anyway.
(btw, I found the software I was talking about! It might help?)

)

The goal is to raise the man's voice (worker) as close as possible to the level of the woman (my mother), while keeping the noise down to an intelligible level.
So this is clearly not meant to sound 'pretty' at all, just understandable because as it is now, there are parts of the recording where you can't hear at all what the guy is saying so you then raise the volume to the max and when my Mom speaks, it will blow your speakers!
Thanks again and of course, anybody else who's up to the challenge is of course welcome to try their skills ;)

wetransfer: https://we.tl/xutD5TKmbm
(knowing my dad, if it really helped, he'll want to tip the kind soul )

 

So you just want to raise the volume of one of the parties in a conversation? No good way to do that in just a couple clicks, but it's a pretty simple process.

Just load the audio file into a DAW, make two audio tracks and chop up the audio file so each "part" of the conversation is its own little segment. Then move all the other guys "segments" onto a separate track and raise the volume.

If you want to clean it up so the ambient noise levels are simliar, that would take a little more effort and knowledge. Like others have said, you could use some kind of spectral editor to isolate the noise and raise the levels on the part that you didn't boost, or remove some noise from the part you raised (although some of those denoisers need a soft touch. they can make things sound a little fucky if improperly used). You'd probably want some crossfading too, to make sure the transitions aren't too jarring.

 

Clip-gain, volume automation, compression, split up voices to different tracks, etc. There are multiple ways.
Denoising with iZotope RX6.

 

Cav Emp, yes, I guess if there really is no easier way, that's the way it will have to be done but right now, I'm still hoping for a better solution because with my skills, on a 30 minute recording, it's going to take me quite a few hours.
Satai kindly offered to give it a shot and I assume he could get it done A LOT faster than I could...so I'll wait and see what comes out of that before I commit anymore time. I've already spent a good amount of time on it, mainly reading manuals and watching tutorials and results are not at all convincing so...

 

Adorable. Here's the result from my crack at it. I improved intelligibility of all the low voices with spectral expansion, adjusted stereobalance, did some overall gentle denoise, and at the end I accidentally put it through some "Beatles" gear chains I was experimenting with on an unrelated project. It sounded so good I left it in. Enjoy!

 

Last edited: Jan 9, 2018

it was a few clicks for me when I help a guy out with the same phone call problem.

 

That's fantastic Satai! Awesome, now we can really hear the full conversation. Thanks so much for all your efforts, it's really very kind of you!
I'll now send it to my dad!

 

Awwww! I love happy endings
Nice work @Satai!

 

well done @Satai , generous of you to apply your skills

I like how he requests English, but she soon goes back to chewing him
out in French again (sounds like )

 

Hello.
Another way to solve this problem would be:
1.Insert the original track in Daw.
2. Apply a gate or a graphic dynamics processor so that only the loud voice is heard.
3. Export the result to file LOUD VOICE, in the same format as the original.
4. Start new Daw session Insert in Daw the original track on track 1, and LOUD VOICE track on track 2. Reverse the phase of only one of the two tracks (180°), with a convenient plugin. That should mute the loud voice on the original track. Only low volume voice should be heard.
5.Export the result to LOW VOICE track, in the same format as original (obviously).
6. Start new Daw session, insert the LOUD VOICE in track 1 and LOW VOICE in track 2.
7. Mix them according to taste, export to EXTRAORDINARY RESULT track.
8. Done. Enjoy.

 

Last edited: Jan 10, 2018