Software to edit music tracks

[jnelsoninjax]

I am curious if such a thing exists, but this is what I am looking for: A program that would allow me to edit a music track and show me all the different channels, real world application would be to load a live track and edit out the crowd noises, clapping, etc. Does such a program exist, and if so would it be something that is exclusive to music producers and be really expensive?

[+BudMan MVC]

What exact source files are you going to be editing?  Your typical mp3 would not have these on different tracks/channels..

 

https://www.audacityteam.org/

 

Is some great free software that can be used to do some amazing things.. But if you think your audio file is just going to have a track/channel that is just say the audience sound??

[cacoe]

If you're familiar with layers in photoshop, imagine all audio tracks as layers that have been flattened. If you want to get part of the image seperately and you only have the flattened layer, because it's now for example been saved as a png and you don't have the PSD, then you just have to cut it out yourself. This can be much more difficult with audio because many sounds exist in overlapping frequencies of other sounds.

 

Best bet is probably Adobe audition -

 

The bottom half of the screenshot, that tool is normally for cleaning up sound, but you can actually select certain regions of sound and past it into a new file.

 

Problem is, this would be extremely difficult to do well, and extremely difficult to do for all of the tracks in a file.

[jnelsoninjax]

12 minutes ago, BudMan said:

What exact source files are you going to be editing?  Your typical mp3 would not have these on different tracks/channels..

 

https://www.audacityteam.org/

 

Is some great free software that can be used to do some amazing things.. But if you think your audio file is just going to have a track/channel that is just say the audience sound??

The files would be lossless like WAV, I can always convert them to MP3 later. I'll look into Audacity, thanks!

[Dick Montage]

As I read this, your question is "Can I edit out certain parts of a sound?"

 

Audio is stored in many ways, and to edit out "parts" you need to understand what "parts" are:

Channels (left / right) - different instruments tend to be placed in different areas of the stereo separation.  Guitars and pianos are often to the left or right, vocals towards the middle.

Frequencies - different instruments produce different frequencies of sound.  Bass drums are lower frequency than a screeching guitar solo

 

So, for example when software tries to remove vocals, it looks for relevant frequencies towards the middle of the audio spectrum and filters them via a gate.  It's rarely perfect as it overlaps other sounds.  Drums tend to be central, for example, and the frequency range of vocals tends to include those produced by guitars (to an extent).  So you end up with the part of the sound that was the vocals more "muddied" than "removed".

 

The analogy made above of layers is a good one to represent individual instrument elements.  Say you have a full band performing a song, you'd have dozens of layers (a drum kit alone can be 20). The correct phrase for these when wanting to get them is "stems".

 

I believe you are asking about removing crowd noise from a live recording.  So sure, you can fade in/out the start/end of the track to remove some of that.  But if you want to remove people clapping along with a song - I mean, you could isolate the frequency of the claps, find where they are in the stereo spectrum and filter it out but... It's gonna sound terrible.

[DevTech]

On 6/3/2019 at 6:58 AM, jnelsoninjax said:

The files would be lossless like WAV, I can always convert them to MP3 later. I'll look into Audacity, thanks!

 

On 6/3/2019 at 7:52 AM, Human.Online said:

I believe you are asking about removing crowd noise from a live recording.  So sure, you can fade in/out the start/end of the track to remove some of that.  But if you want to remove people clapping along with a song - I mean, you could isolate the frequency of the claps, find where they are in the stereo spectrum and filter it out but... It's gonna sound terrible.

The word he is looking for is FILTER. All sorts of filters are available as Plug-ins and most audio editing programs can use these plug-in filters.

 

If you have control of the original recording, then you can improve your chances by using more than a stereo pair recording. If it is a musical performance, then each performer could have a directional mic pointed at them and for filtering you could point a mic at the loudest interference sources so you can invert that signal later.

 

Would still need to feed that to a "smart" filter which is why you mentioned expensive professional stuff which I'm sure exists. This scenario is perfect for some AI type ML processing and if that exists you might find some free research projects on GitHub...

 

[exotoxic]

Don't waste your time. If you don't have the source files all you can do is filter which will ruin your audio. The only hope would be if the stuff to filter was spaced out enough to not overlap the music.

Edited June 8, 2019 by exotoxic

[Dick Montage]

19 hours ago, DevTech said:

 

The word he is looking for is FILTER. All sorts of filters are available as Plug-ins and most audio editing programs can use these plug-in filters.

 

If you have control of the original recording, then you can improve your chances by using more than a stereo pair recording. If it is a musical performance, then each performer could have a directional mic pointed at them and for filtering you could point a mic at the loudest interference sources so you can invert that signal later.

 

Would still need to feed that to a "smart" filter which is why you mentioned expensive professional stuff which I'm sure exists. This scenario is perfect for some AI type ML processing and if that exists you might find some free research projects on GitHub...

 

I feel you missed my point. There is no such filter that works perfectly on studio recorded audio, let alone a live concert recording that has a crowd.

 

All filters can do is remove, and when doing so they ALWAYS remove more than needed because it’s always a best guess.

[devobitch]

https://www.celemony.com/en/melodyne/what-is-melodyne

 

[DevTech]

2 hours ago, Human.Online said:

I feel you missed my point. There is no such filter that works perfectly on studio recorded audio, let alone a live concert recording that has a crowd.

 

All filters can do is remove, and when doing so they ALWAYS remove more than needed because it’s always a best guess.

There was no chance he would realize the HUGE variety of filters available.

 

The implication is that he is the person doing the recording and so he can add many more mic inputs if he chooses to greatly improve filtering options

 

For example a directional mic on the clapping recorded at the same time of course could be inverted and filtered in to achieve much better accuracy. Enough? who knows. And even that is primitive:

 

I did a survey a while back of research papers on GitHub for processing images with AI and the progress has been fantastic. I would assume that similar progress might be possible in the audio domain, but he can do the searching if he wants, It's probably there...

 

 

 

[Dick Montage]

Sorry I think you’re misreading and assuming way too much. But ok, I’ve given my answer and I’m not here to argue. My many years of music production tell me all I need to know here, having worked with digital audio since the early 90s.

[DevTech]

3 hours ago, Human.Online said:

Sorry I think you’re misreading and assuming way too much. But ok, I’ve given my answer and I’m not here to argue. My many years of music production tell me all I need to know here, having worked with digital audio since the early 90s.

The OP made a guess that some sort of high end filter might exist and be very expensive.

 

I was just saying there might be a case for that guessed scenario and that it might involve more than a single L/R mic pair so like a noise cancelling array, there is more info for the algorithm.

 

But of course an analog frequency bandpass filter is not going to do it or even a finely controlled digital equivalent. It would probably involve a machine learning algorithm similar to how they do face and body tracking in video and substitute other entires. If you train an algorithm on human clapping and then add a parameter tuning algorithm via Reinforcement Learning you might be able to do it.

 

Whether there is any real value to that type of alteration beats me. Maybe add a "slow clap" sound to that doctored Nancy Pelosi video that went viral in alt-right circles...

 

Or what is the sound of one hand clapping? If we edit that out, nobody will know...

 

 

EDIT:  When I suggested that the OP search out some research papers, I thought that would make it clear that there is no off the shelf tech that I know of that would filter out clapping. I apologize to anyone in this thread who might have got a different impression from my thoughtful musing on the subject.

[DevTech]

Maybe it is useful to get up to speed with how audio processing can be done outside the limitations of twiddling knobs. Here is a survey of some of the ways AI is applied to audio processing:

 

https://github.com/ybayle/awesome-deep-learning-music

 

And here is an example of separating out the vocals in music which I'm guessing is harder than clapping:

 

https://github.com/andabi/music-source-separation

 

 

A demo of an earlier version of that algorithm:

 

 

 

 

 

 

 

[Dick Montage]

Look, you're trying and condescending, but still wrong.  But that's ok.  The removal of crowd noise is MUCH harder than isolating studio vocals.  Twiddling knobs, OK sure.

[DevTech]

1 minute ago, Human.Online said:

Look, you're trying and condescending, but still wrong.  But that's ok.  The removal of crowd noise is MUCH harder than isolating studio vocals.  Twiddling knobs, OK sure.

You are going out of your way to be offended by my serious attempt to wonder if the objective of the OP is even possible.

 

Very sophisticated processing is starting to be applied to audio, just like it has been applied to video and we are seeing nascent results that look promising.

 

"Twiddling knobs" is a common phrase to refer to varying of parameters on things like parametric equalizers and are limited by an overall approach based on analog thinking.

 

By using algorithms that produce processing that humans can no longer understand, huge advances in the computing industry have been happening in the last few years. There is NO reason to expect that the audio domain will be exempt from this approach!

 

 

 

 

[Dick Montage]

4 minutes ago, DevTech said:

You are going out of your way to be offended by my serious attempt to wonder if the objective of the OP is even possible.

I'll take this at face value and say that I have read offence where none was meant then.  I have been a knob twiddler since the 80s and your "maybe it would be useful to get up to speed" implies that I'm not.  I am.

 

Yes: It absolutely IS possible ON PAPER.  But even the highest level of AI-integrated gating (filtering) and potential reconstruction of the over-filtered sound created by the gates is nowhere near there yet.

 

Let me make another analogy:  Compression can be lossy, and yet still recreate an acceptable replication of the original content.  ANY filtering is inherently lossy, and the acceptability of the output is obviously for the user to determine, but it's never reached a level (and I've been in the studio with number 1 selling artists and producers) that I would consider "good enough".