Tools in Logic to filter instruments?

[GeneralDisarray]

In the interest of learning film orchestration, let's say you're trying to pull apart a song from your favorite composer. Some songs are easy, being just one violin or cello, or something like that, with a clear melody in legato from start to finish. Those are the easy ones, and there's nothing wrong with that since many times they are also beatiful.

But many others have different section of violins playing the same note progression at different octaves, or totally different notes at the same time, so it gets a bit confusing. It gets even more confusing where it's the same instrument playing different note progressions. Sometimes they're not so hard if you enable the loop in Logic and you listen to it over and over about 50 times, your brain starts to separate them, even better of course if the two sections are panned differently.

And if it's bass it's kind of difficult for some of the lowest notes, but at the same time, easier because you just drag the last band in the EQ and you remove basically everything else. Same goes if there's a melody playing that is very high pitched, many times (but not always because high notes don't always mean very high frequencies), you remove everything below, let's say, 4 Khz.

But for those that are in the middle like I mentioned before, are there are other tools in Logic, or maybe 3rd party that are not crazy expensive, that will allow me to isolate a violin section from another? And I don't mean in a way that will sound good at all, just a way that will tell me the note progression basically.

[zplane]

It could help to use the flex/varispeed option in Logic to slow down the tempo of the audio so you can hear what's going on in all the parts.   We don't recommend students do this when we assign them recordings of music to transcribe, but it often helps them get started.  (We can't really stop them from doing this except on in class transcription tests!)    If the arrangement is really dense though it can help.   I don't know what your musical background is, but if you can read music, you could get a copy of the orchestral score.  Following the score during playback and seeing the notes for the parts you can easily hear should help you identify the notes for the other parts that may not stand out in the mix.  That can help focus your ear on hearing the notes your eyes have told you are being played.

Conventional EQ and filtering will only get you so far in trying to separate out the parts.   So vendors have made spectral separation tools to help split the audio file of a complete mix into different frequency ranges and parts.  I believe some recognize the temporal and spectral characteristics unique to the human voice, piano and percussion.   These tools are often designed to take apart final mixes of popular music arrangements and are setup to automatically create separate audio files ("stems") for bass, piano, drums and vocal parts.  Then any other remaining audio in the spectrum is dumped into a single file, e.g., named "other".  These tools cannot isolate say a guitar part from a keyboard/synth part because the frequency ranges overlap.  I have not tried them on a classical recording, but you might get a useful separation between the parts of the orchestra carrying the bass parts compared to the treble parts in the strings.  Steinberg's Spectral Layers separation software has a trial version you could download and experiment with, https://www.steinberg.net/spectralayers/   You might want to try it on a pop song first just to see how it normally works before trying it on something classical.  There will be some audible artifacts in the separate stem files.

 

[fuzzfilth]

41 minutes ago, zplane said:

We don't recommend students do this

Why not ?

[Mania]

Melodyne plugin(have to purchase ) might help you with this. If i remember well it can analyze an entire song and you can change the pitch in certain elements/instruments in the song separately. Might be able to solo/mute stuff too. Haven’t checked lately. It would be in flex pitch style. A quick youtube search would probably reveal what’s possible with Molodyne these days. 

[GeneralDisarray]

19 hours ago, zplane said:

It could help to use the flex/varispeed option in Logic to slow down the tempo of the audio so you can hear what's going on in all the parts.

Thanks, that's actually something that I had tried, although mostly when I wanted to figure out the note progression on a very fast violin, and it works pretty well. But I hadn't thought about using it also for this, so thanks, I'll definitely give it a try.

19 hours ago, zplane said:

Conventional EQ and filtering will only get you so far in trying to separate out the parts.   So vendors have made spectral separation tools to help split the audio file of a complete mix into different frequency ranges and parts.

Yes, actually I had tried one of those, I can't remember the name now but it had been recommended by Venus Theory in one of his videos. I found it remarkable that you could extract so much info from an audio track. However, the program was really sluggish, and when your program is sluggish in one of the fastest Macs ever made, then you have to go back to the board, at least when it comes to graphics. And besides that, it was really weird, and one of the least intuitive programs I had ever used. Plus it wasn't as expensive as Melodyne, but it was still quite expensive, so I ended up not buying it.

But I figured maybe Logic has something more basic than that which could at least get me going. I just didn't mention it in my post because I still can't remember the name of the program, and even less that the technology is called spectral separation.

I had seen a video months ago about how to extract simple note progressions in Logic using Flex/Varispeed, but I tried that last week and it didn't work, because there was one step where I was supposed to choose a menu item and in my case it was grayed out. Regardless, to me what's fun and educational is to listen to a bar or two over and over and guess what note is first, then second and so on, and that's also training my brain to distinguish notes to the point that hopefully some day in the future I can hear a note and say "That's a B2".

[zplane]

The spectral separation tools I've used are Steinberg's Spectral layers pro (mentioned above, https://www.steinberg.net/spectralayers/trial/).  Then Hit'nMix's Ripx, https://hitnmix.com/ which some users report is better than Spectral Layers for trying to fine tune the separations.  I would agree that spectral editing in these is not exactly intuitive.

Mania's suggestion about Melodyne is a good one.  I think you would need the version that supports polyphony or separating out notes from multiple parts in the same audio file.  That would be the Melodyne Editor version, https://shop.celemony.com/cgi-bin/WebObjects/CelemonyShop.woa/wa/shopEntry?task=displayDetail&article=510&country=US.   It looks to me like the 30 day trial version has all the advanced functions of the Editor version, https://www.celemony.com/en/trial, so you could try that.

To follow up on fuzzfilth's comment, for people playing in ensembles and groups, ultimately we want them to be able to identify pitches and develop relative pitch recognition of intervals while they are playing music on their instrument at the correct tempo.  When you slow things down your ear/brain auditory system has more time to lock on to and identify the tonal center of a pitch.  For some people the speed at which they recognize pitches and intervals improves with practice and their brain starts to require less time to process and recognize pitch information.

[fuzzfilth]

1 hour ago, zplane said:

ultimately we want them to be able to identify pitches ... at the correct tempo

I thought as much. But then, doing it slowly first seems to be the general consensus in all matters between learning a skill and making love.