tmc
Wholenote
Posts: 893
Formerly Known As: tmc
|
Post by tmc on May 22, 2021 11:23:57 GMT -5
I've been watching YouTube videos where they have managed to isolate the individual instruments. Is it possible to get this level of isolation by applying digital filters to eliminate everything except the target instrument?
Keith Richards isolated bass line on Sympathy For The Devil.
|
|
|
Post by FlyonNylon on May 22, 2021 19:30:44 GMT -5
That’s pretty isolated. It would be very difficult to get it that good with a mid-range instrument like guitar/piano/voice because there’s so much else in the mix at those frequencies.
EQ and gates can do a lot. I’m sure theres a plug-in specifically for frequency isolation but I really doubt they could get the accompanying piano part in that song as isolated. Who knows though.
|
|
|
Post by Peegoo 🏁 on May 22, 2021 20:54:43 GMT -5
Pulling out individual instruments or vocals from a stereo mix that cleanly is impossible. What you're hearing on YT are called "stems" taken from the original recordings.
These are the discrete individual tracks extracted from the studio multi-track recording. Up until about 15-20 years ago almost every studio recorded on 24-track Studer or TEAC/Tascam tape decks. These days it's all computerized, but there are still discrete tracks for each instrument (sometimes more than one with various effects/treatments applied in real time).
Record producers know each other and they often share these stems with each other. There are a lot of them floating around these days, and some make it to the Innerwebz.
|
|
|
Post by Peegoo 🏁 on May 22, 2021 21:05:08 GMT -5
Also, when you see the words Digitally Remastered, it means the original multitrack tape recording has been migrated to the digital realm as individual tracks. Next, the digital version of each track is..."tweezed" to perfection (that is a relative term, depending on your musical tastes), and then all individual tracks are remixed down to a stereo master.
|
|
tmc
Wholenote
Posts: 893
Formerly Known As: tmc
|
Post by tmc on May 22, 2021 21:36:28 GMT -5
I was wondering if software used in digital instruments, emulating physical instruments, could be used as a mask, to filter non wanted frequencies and overtones. The human ear, (not mine), can discern vocals, guitar bass, drums, etc. Maybe an AI algorithm could be used as a mask to isolate a targeted instrument or voice.
|
|
|
Post by FlyonNylon on May 22, 2021 22:23:53 GMT -5
first listen was on my phone. listening again in the studio and watching more from the creator, that's definitely just the stems not mixing/plugins! lol.
The answer to your question is complex. While typically vocals, bass, and drums occupy different frequencies, there are overtones in each part that fill the spectrum. So doing just a low-pass filter to isolate bass makes the sub/low frequencies more prominent but reduces the overtones and the overall effect of the instrument. It's a trade off. Mixing drums you can use gates to allow specific frequencies to be amplified but using hard gates on instruments causes an uneven sound.
In answer to your question, yes, mixing in a modern DAW can filter unwanted frequencies and overtones. It just can't eliminate them from a stereo-mix because of everything else going on.
|
|
pdf64
Wholenote
Posts: 553
Member is Online
|
Post by pdf64 on May 23, 2021 4:09:52 GMT -5
Thanks, I’ve not come across the ‘stem’ terminology before
|
|
|
Post by Auf Kiltre on May 23, 2021 7:47:26 GMT -5
A while back I was listening to some isolated tracks of Levi Stubbs from The Four Tops and James Jamerson on bass doing Bernadette. I thought the old Motown stuff was done on 3 track, so I could see the vocals on a dedicated track but the bass was a puzzler. Both tracks had bleed, but if that was digital isolation technology it's come a lot farther than I ever thought it would. I remember the earlier days on the Vegas boards (I cut my teeth on Sonic Foundry/Sony Vegas) and the iso topic was a frequent newb topic that would get the beenie propellers twirling with the techies ("It can't be done!!").
|
|
|
Post by Peegoo 🏁 on May 23, 2021 10:55:44 GMT -5
A picture makes this easier to understand. Light is similar to sound; each color has a specific 'center' frequency that does not change. If you apply a filter to isolate the blue color from sunlight, you will either pull in some indigo and some green with it by using a broad filter like this, or you will omit some of the blue by using too narrow a filter, like this. Sound is complicated by the fact that most instruments do not occupy their own individual space in the sonic spectrum the way light does; there is quite a bit of overlap. You can use parametric EQ to pass a specific frequency range within a stereo master, but it cannot completely remove the sounds of other instruments that appear in the same range. This sort of stiff filtering results in fairly unnatural instrument sounds because much of the range of the instrument is filtered out. Want to have some fun? You can easily make yourself a little device that allows you to listen to familiar stereo recordings in a whole new way due to phase cancellation. With some recordings--the effect is minimal...but with others, it can be quite dramatic. Sometimes the vocals (or other tracks) are almost completely gone. Stereo headphones use a TRS (tip-ring-sleeve) plug with three wires and are configured to maintain left/right channel separation. You can make yourself a little patch cable with a scrap TRS plug and lead, e.g., clipped off of old/broken headphones, and two little alligator clips. Omit the sleeve conductor in the lead, and make sure the ring clip makes contact with the ring AND sleeve on your headphone cable's plug so you'll hear sound through both ears. You cannot hurt your equipment doing this...just don't short the alligator clips to each other. Most headphone cables's color codes are red (tip), white (ring), and black (sleeve). Test your patch cable first for continuity to confirm this before you snip the black wire too short. Plug the little patch cable into your music device, clip the other end to your headphone plug, and you're off to the races. If you try this, report back and tell us what you think!
|
|
|
Post by Auf Kiltre on May 23, 2021 11:20:35 GMT -5
Phase cancellation is an interesting phenomenon. I'll commonly use the brainworx bx_solo stereo widener plugin on my mixes. It does exactly what it's suppose to do, widen the stereo spectrum. I never go past 150%. But every now in then it has adverse effects on tracks, like intensifying the reverb on say a vocal track, or actually making something like a guitar track that has stereo delay/time based effects on it inarticulate or even seem to disappear on some sound systems.
I remember way back one of my wife's nutty friends got a home karaoke system that even then used some sort of phase cancellation to allegedly strip the lead vocal for the wannabe Siouxsee and the Banshees karaokiest. What usually resulted was a washed mess that sounded like the lead vocal buss was turned off but the send to the effects channel was still turned on.
|
|
|
Post by ninworks on May 23, 2021 11:27:30 GMT -5
A while back I was listening to some isolated tracks of Levi Stubbs from The Four Tops and James Jamerson on bass doing Bernadette. I thought the old Motown stuff was done on 3 track, so I could see the vocals on a dedicated track but the bass was a puzzler. Both tracks had bleed, but if that was digital isolation technology it's come a lot farther than I ever thought it would. I remember the earlier days on the Vegas boards (I cut my teeth on Sonic Foundry/Sony Vegas) and the iso topic was a frequent newb topic that would get the beenie propellers twirling with the techies ("It can't be done!!"). Some of the recording techniques they did with the primitive 3 and 4 track machines involved using multiple master multi-track tapes. If it was a 3 track they could record say, drums, bass, and guitar, on the first 3 tracks and then mix those 3 tracks to combine them and record them onto 1 track on another 3 track machine. That would leave 2 tracks available for overdubs on the new tape. Then they would do it again to yet another tape leaving 2 more tracks available for overdubs. Vocals were usually recorded last so the generation losses would be the least and the vocals would sound good. They didn't usually go more than a generation or two because the losses were too severe. I know the original 4 track recorders used a 1" tape width. I think the 3 tracks were also 1" tape but I'm not certain of that. Spreading 3 or 4 tracks across a 1" tape makes for a very wide track width so the dynamic range and audio quality of those machines was ridiculously good. Theoretically much better than the 2" 16 and 24 track machines of later years. With digital technology they can go back to the first original 3 track tape and record them to 3 tracks digitally. Then they can take the 2 new overdubbed tracks from the second tape and record them onto 2 more digital tracks. They can do that as many times as needed to get the original takes from the first generation multi-track masters so the digital multi-track has all the original takes without any generation losses. Then they can sweeten and remix the original first generation tracks digitally and master that. That's why many of the early, remastered, Beatles, albums sound so much better than the originals. All the recorded tape tracks are first generation so the frequency spectrum and dynamic range losses are minimized. Digital isolation technology has come a long way but with complex program material such as a multi-instrumental mix it can't pick any one thing out and isolate it completely. That's not possible yet. There are some digital forensic techniques that can be used to help but it's not 100%.
|
|
|
Post by ninworks on May 23, 2021 11:31:56 GMT -5
Phase cancellation is an interesting phenomenon. I'll commonly use the brainworx bx_solo stereo widener plugin on my mixes. It does exactly what it's suppose to do, widen the stereo spectrum. I never go past 150%. But every now in then it has adverse effects on tracks, like intensifying the reverb on say a vocal track, or actually making something like a guitar track that has stereo delay/time based effects on it inarticulate or even seem to disappear on some sound systems. I remember way back one of my wife's nutty friends got a home karaoke system that even then used some sort of phase cancellation to allegedly strip the lead vocal for the wannabe Siouxsee and the Banshees karaokiest. What usually resulted was a washed mess that sounded like the lead vocal buss was turned off but the send to the effects channel was still turned on. I used to put up a 3rd, rear, speaker on my stereo back in my yute. You would wire the positive wire from each of the stereo channels to another speaker and it would cancel out anything that was mono in the mix leaving only the things panned right and left coming through. Pseudo-surround sound on a budget. I even wired in a potentiometer to turn it down because it was usually too loud.
|
|
|
Post by Auf Kiltre on May 23, 2021 11:50:06 GMT -5
Ahh yes, I forgot the possibility of the machine to machine transfers. Heck, I did that in the early days with 2 cassette units, one which had the ability record both line and mic sources at the same time.
I was listening to "My Girl" through cans the other day and realized the harder drum hits sounded distorted. I don't know if that's an initial source problem or a result of transfers.
|
|
|
Post by Peegoo 🏁 on May 23, 2021 11:52:26 GMT -5
You would wire the positive wire from each of the stereo channels to another speaker and it would cancel out anything that was mono in the mix leaving only the things panned right and left coming through. That is exactly what that little TRS cable hack does--it cancels out much of the content that's at or near 12 o'clock in the stereo mix.
|
|
tmc
Wholenote
Posts: 893
Formerly Known As: tmc
|
Post by tmc on May 24, 2021 10:39:55 GMT -5
I tried the TRS hack on Creedence's Ramble Tamble. The vocals almost disappeared leaving an interesting mix of drums, bass, and guitar. I enjoyed that.
|
|
|
Post by oldfartbassplayrwalt on May 24, 2021 20:59:53 GMT -5
I can't find the article, but somewhere was a description of how the Beatles' "hollywood bowl live" tracks were completely rebuilt in the last two or three years.
Something along the lines of Peegoo's explanation, but using AI software to listen 'only' for the frequency profile of each instrument, and then extract and isolate those sounds, then 'Remix'ing the individual tracks as if it was a real time capture.
Kind of scary technology, if I got the explanation right...
|
|