Splitting a track into frequencies

Is it possible to split a track - any track, but say a vocal track - into two, or three frequency bands, and then place those bands in various place across the stereo width, or even put them on separate tracks? Is there a plugin to do that?
Thanks
 
Perfectly possible but what on earth would you do it for? It would sound weird and imagine the stereo field wildly thrashing about. A few people record pianos very wide and it sound really strange, I think a voice with just one note at a time might be a great special effect. I’ve never seen a plug in as it’s such a strange thing, but you could do it by duplicating the track and just panning each eq’d band to its own pan position.
 
Perfectly possible but what on earth would you do it for? It would sound weird and imagine the stereo field wildly thrashing about. A few people record pianos very wide and it sound really strange, I think a voice with just one note at a time might be a great special effect. I’ve never seen a plug in as it’s such a strange thing, but you could do it by duplicating the track and just panning each eq’d band to its own pan position.

Thanks. My friends think I'm weird too, well..................
 
Well, I have been making copies of my bass tracks to treat each one different with EQ and compression. Example, one track I EQ HP filter and compress it one way. Then another with LP and compress it another way. Then group them together for a final out with smaller amount of compression.

I can see a reason for it, even if it is crazy ;)
 
Just duplicate (or triplicate) your track and use your EQ plugin. You can put a highpass, lowpass and both on each with the frequency and slopes set so the mix is more or less the original track. Keep a copy of the untouched track and A/B against a mono version of your 2 or 3 track mix is what I'd do.
 
The shift to one side and back on a bass part would be quite strange though wouldn't it - but maybe it would be great on old 70s disco - Earth, wind and Fire etc - octaves one side, lowest the other - that could be great on headphones? Maybe I'll try it and see?
 
How about Boogie Wonderland - three copies of the same track then each one split low, mid and high. I had trouble with the mid one - difficult to find more than a handful of notes if I went too narrow, so the middle track is quite low level.
 

Attachments

  • bass.mp3
    9.8 MB · Views: 7
Just duplicate (or triplicate) your track and use your EQ plugin. You can put a highpass, lowpass and both on each with the frequency and slopes set so the mix is more or less the original track. Keep a copy of the untouched track and A/B against a mono version of your 2 or 3 track mix is what I'd do.

That's the basic solution for your task as stated. (You could also call both a high and low pass together a "band pass"). I use this technique for mastering (not for individual tracks). I like it, but you do have to watch phase.

If you meant you want to split based on fundamental frequencies (i.e. the note being played rather than just the frequencies themselves), life is more difficult. Melodyne and similar tools can split an instrument up into individual notes, at which point you could do whatever panning. It'll be easier with monophonic instruments. Regardless, you might have to do a fair amount of manual editing.
 
Isn't this basically what was done on many early "electronically reprocessed for stereo" albums? (those of you under 55 would probably never have heard of this :) ).

I remember my Dad having lots of albums from the late 50s. It was called Duophonic.
 
Example: take a particular vocal, cut the vocal range in half. put one here and one there.

Yeah, that'll be a pain. I don't know of any tools to fully-automate it, but melodyne will give you some ability to manipulate the individual notes of the vocal track
 
If you were running REAPER, you could:
1. Click the routing button on the track in question and choose 6 TrackChannels
2. Insert the plugin JS: 3 Band Splitter
3. Assign the output pins of the three band splitter to whichever tracks you'd want for the different frequencies. Or even do it all on one track using JSChanmix
 
One thing no one has mentioned - a vocal note is not just ONE note, it has a series of harmonics and overtones. Without them, you would sound like a sine wave. So by using selective EQ to isolate a frequency range, it would no longer sound like a normal voice, because it would be missing the other associated frequencies.
 
Okay, I think I understand what you're after, and I think it's worth looking at, but a couple things to consider. I think the frequency splits would have to be managed as picking a specific frequency would give you different things for different notes, no?
If you set your split at 1000 hz, and split the lo pass left and the hi pass right. You sing C5 @ 523.25 hz, and then sing C6 at 1046.5, the effect on the harmonics and overtones (that Mike pointed out) would be completely different unless you picked a different split point for each note. Otherwise the effect would in effect walk across the stereo field from left to right dependent as the scale increased...which could be cool in and of itself.

If, however, you split at the note center for each note or some interval (OH GOD THE MATHS) up or down, the effect would be uniform.
 
One thing no one has mentioned - a vocal note is not just ONE note, it has a series of harmonics and overtones. Without them, you would sound like a sine wave. So by using selective EQ to isolate a frequency range, it would no longer sound like a normal voice, because it would be missing the other associated frequencies.

Which is why they'd need melodyne or some other suite with magic algorithms to isolate whole notes by fundamental frequency
 
It is!

As a programmer, I find the way people toss around the term AI grating, but some of those "AI-driven" algorithms are pretty amazing!

AI refers to artificial intelligence, meaning an ability to learn. A preprogrammed algorithm is not AI, just smart programming by organic intelligence. :)
 
AI refers to artificial intelligence, meaning an ability to learn. A preprogrammed algorithm is not AI, just smart programming by organic intelligence. :)

That's not really how it's used in industry anymore. Mostly, it just means "the algorithm was trained on a large data set and is in some ways self-writing"
Of course, how your average dev house is going to use it just means "uses Google's API for object recognition in photos" or "uses a standard 'machine learning library' but only with our own small data set"
 
Back
Top