| Author |
 Topic 
|
Lendrik
Location: United Kingdom
Posts: 17
|
 Posted - Sun Jun 29, 2003 6:11 am
|
|
|
First of all, I've searched the forum on this topic and I can see that a lot has been talked about it and that people here are quite anoyed by the whole thing. It's been stated many times that to remove instruments (left & right channels) and keep vocals (centre channel) is not possible. Well, it may not be possible on Cool Edit (at least at the moment) but it is possible in theory. So, listen, maybe it could be implemented in the future version of Cool Edit.
Firstly, here are some basics:
1. A stereo recording has two audio channels - left and right.
2. Some instruments are located on either channel, but vocals, bass and drums are usually located on the virtual "centre channel", meaning they sound equal on both left and right channels, making it sound as if they come from the middle.
3. It is possible to remove that "centre channel", making a vocal-cut version, using a well known method (Cool Edit has a mixing preset for that).
So, what if somebody wants to do opposite, to remove left and right channels, keeping only centre channel (in other words, to isolate vocals)? Well, one of my hobbies has been programming in QBasic, and I guess anybody who's familiar with programming knows about boolean operators - AND, OR, NOT etc. They are used to compare different data to each other. Now, what we need to isolate the centre channel, is a program or a plug-in that would compare left and right channels to each other and leave data that is present on both channels only. That is, any instrument that sounds only on the right (or left) channel would be cancelled out.
I hope this doesn't sound too complicated, it's easy in theory but I don't have enough programming knowledge to implement it in life.
|
|
|
|
zemlin
Location: USA
Posts: 1156
|
|
Posted - Sun Jun 29, 2003 6:24 am
|
|
|
|
When I want to beef up the center I take the results of a vocal cut and subtract them from the original stereo file. It removes the stuff that was leftover from the vocal cut leaving the center channel signal.
|
|
ozpeter
Location: Australia
Posts: 3200
|
|
Posted - Sun Jun 29, 2003 6:42 am
|
|
|
| Quote: |
| As for removing the instruments and keeping the vocals, this cannot be done - if you try to isolate the centre of the stereo image, where the vocal is usually found, you will still have any other sound that was in the centre, plus half of what was left or right of the centre. You can try this using the Cool Edit 'Channel Mixer', first applying the 'LR to Mid Side' preset to your file, then using the 'Both=Left' preset on the result. |
6400 people have read those words and not challenged the veracity of the above quote from the 'sticky thread'!
- Ozpeter
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Sun Jun 29, 2003 7:11 am
|
|
|
| Lendrik wrote: |
| I hope this doesn't sound too complicated, it's easy in theory but I don't have enough programming knowledge to implement it in life. |
.. or enough theoretical knowledge of audio to realise that this is exactly what the CEP preset does - and that because everything else that you mentioned is still dead centre, it remains as well. ozpeter is right - nobody has questioned the accuracy of the sticky thread content because it is a self-evident truth.
Forget programming it - you will end up with an identical result.
_________________
 |
|
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Sun Jun 29, 2003 7:15 am
|
|
|
| Lendrik wrote: |
| ....it's easy in theory but I don't have enough programming knowledge to implement it in life. |
Unfortunately, it's most certainly not easy. In fact, it's not possible, as you will find stated on this forum on more than one occasion. I'm afraid it's because you have failed to understand the underlying principles and are postulating a theory based on this misunderstanding.
You'll also find some in-depth discussion of these fundamentals oin the archives, so I'm not about to go through all that again. If you are really interested, you'll have to go and look for it yourself. ISTR SteveG has made some lucid explanations of the real theory invoved.
However, let's take the philosophical approach and look at it another way (why not, it worked well for the Greeks);
1 - it's obvious there is a demand for this function (although, I have to confess, I'm not sure why  ). From this, it follows there is a potential market for such a product.
2 - the world is full of programmers who, unlike you, do have the knowledge required to implement anything you can think of (if it is theoretically possible to do, that is).
3 - search all you will, but you will not find software or hardware - at any price - which does the job well.
Ergo - it can't be done.
QED
|
|
ozpeter
Location: Australia
Posts: 3200
|
|
Posted - Sun Jun 29, 2003 7:17 am
|
|
|
And the 'MS Mixer' add-on for Cool Edit remains available from me upon email request, for those who want to fiddle with M and S balances and eq the easy way. (Karl?!).
- Ozpeter
|
|
William Rose
Location: USA
Posts: 467
|
|
Posted - Sun Jun 29, 2003 6:05 pm
|
|
|
I, being one of Ozpeter's MSMixer customers, can report good results from it's use. It does as good a job as anything can, I think.
Graeme, you really can't imagine why anybody might want to do this ?
Speaking for myself, it can be a fabulous tool, along with CEP's "Stretch" function, for learning guitar parts. Especially during verses, where someone starts a syllable at the same instant a note is struck, or anywhere vocalization obscures a quieter instrument.
No, the vocals aren't gone of course, but for my purposes they don't have to be. They just have to be a little less dominant.
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Mon Jun 30, 2003 6:26 am
|
|
|
| William Rose wrote: |
| Graeme, you really can't imagine why anybody might want to do this ? |
Perhaps I was overstating my case, because I accept there are a few things which one might do where it would be a useful thing to be able to do this.
You gave a good example of one, but I strongly suspect that the majority of people asking about it are just too damn lazy to learn how to play an instrument in the first place. It's all part of the 'one button', instant gratification society in which we live.
|
|
MusicConductor
Location: USA
Posts: 1524
|
|
Posted - Mon Jun 30, 2003 9:43 am
|
|
|
|
Aw, c'mon Graeme, you mean I have to sing or play something if I want to make a recording? I mean, how archaic is that?
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Mon Jun 30, 2003 9:52 am
|
|
|
| MusicConductor wrote: |
| Aw, c'mon Graeme, you mean I have to sing or play something if I want to make a recording? I mean, how archaic is that? |
:D
|
|
HanzZ
Location: Netherlands
Posts: 334
|
|
Posted - Mon Jun 30, 2003 9:56 am
|
|
|
I used the vocal cut option last week. Want to know what happened? It took out my appendix and trimmed my toenails... Man, threethousand years in showbusiness and I never handle such a routine like that...
-- HanzZ
_________________
I make music like apples fall from trees... |
|
|
|
rfchmbrs
Location: USA
Posts: 30
|
|
Posted - Mon Jun 30, 2003 11:54 am
|
|
|
I'll disagree with the comments about impossible to separate the noise (vocal) from the signal (music). To me the question is a matter of cost and computer time. There are other areas of sound recording that I know use methods that are applicable. Biggest problem is cpu time and people time  . These methods would need to be trained at the outset of a project and then interacted with as the process continues. The end product would be the separate noise file, signal file and an error file.
I plan to use CEP as a front-end for just such a tool and plan to start coding in about 3 months. I have other projects on the table right now and they are for pay  .
RON C 
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jun 30, 2003 12:00 pm
|
|
|
| rfchmbrs wrote: |
I'll disagree with the comments about impossible to separate the noise (vocal) from the signal (music). To me the question is a matter of cost and computer time. There are other areas of sound recording that I know use methods that are applicable. Biggest problem is cpu time and people time . These methods would need to be trained at the outset of a project and then interacted with as the process continues. The end product would be the separate noise file, signal file and an error file.
I plan to use CEP as a front-end for just such a tool and plan to start coding in about 3 months. I have other projects on the table right now and they are for pay .
RON C |
I'm glad you've got the time to waste...
_________________
|
|
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Mon Jun 30, 2003 2:27 pm
|
|
|
| rfchmbrs wrote: |
| I'll disagree with the comments about impossible to separate the noise (vocal) from the signal (music)..... I plan to use CEP as a front-end for just such a tool and plan to start coding in about 3 months. |
Hohoho - ROTFLMAO. I think we've been here before.
I seem to recall another member who was going to code a 'mic modeler' which would actually model microphones (which current software can not do). There was a long discussion about the whole subject and he finally realised that it was nowhere near as easy as he had first thought, but was still convinced it could be done (it can't, by the way). Needless to say, some of us disagreed with him, but we offered to beta test anything he might come up with.
We've never heard a word about this project since. I presume the reality finally struck home.
So - I'm making my offer now. You code, I'll beta test it (and I expect a few other guys here would make the same offer).
Now, I'll sit back until reality strikes home for you. At least it should hasten the end of this thread.
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Mon Jun 30, 2003 2:48 pm
|
|
|
I just want to make one thing clear - this topic is not about vocals vs instruments, it's about extracting the centre channel. In other words, keeping only those sounds which appear equally on both left and right channels. That of course includes vocals along with bass guitar and bass drum, but the last two can be filtered out later.
With the centre channel extracted, it can be then inverted and mixed with the left and right channels of the recording to get the "pure" left and right channels, that is, without the vocals and bass, but still a stereo track, unlike the usual "vocal cut" track, which is a mono track with either left or right channel inverted and mixed together. Phew, that was a long sentence to write 
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jun 30, 2003 3:20 pm
|
|
|
| Lendrik wrote: |
With the centre channel extracted, it can be then inverted and mixed with the left and right channels of the recording to get the "pure" left and right channels, that is, without the vocals and bass, but still a stereo track, unlike the usual "vocal cut" track, which is a mono track with either left or right channel inverted and mixed together. Phew, that was a long sentence to write |
And it was wrong! CEP vocal cut does exactly what you describe your version to be. If you do an analysis of a stereo signal, you will find that it can be rearranged into a sum and difference signal. The sum is mono, and the difference is the stereo information from both channels - ie, whatever isn't in the mono channel. Inversion does not enter into this, but subtraction does. The sum and difference equations are well-understood, and cannot be disputed, even by you.
And if you don't understand even this much, then you are doomed to fail, because there is an absolute prerequisite for being able to code anything - and that is that you understand the problem and issues that it raises in the first place. What amazes me is that you don't even seem to have studied how the CEP vocal cut works, and you are mouthing off about it as if you are the only person in the world who has ever thought of 'your' method. You are not. Not only that, but you don't seem to have anything like the right attitude towards investigating anything.
_________________
|
|
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Mon Jun 30, 2003 3:44 pm
|
|
|
| Lendrik wrote: |
| With the centre channel extracted.... |
What centre channel? There is no centre channel - it's an illusion!
OK, I know we talk of it, but we are aware of the fact that it doesn't really exist and the term is conventinally used for convenience's sake.
There are two channels one left, one right.
That's it!!
No more than two,
no centre,
no way can you make it work.
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Mon Jun 30, 2003 3:45 pm
|
|
|
| SteveG wrote: |
And it was wrong! CEP vocal cut does exactly what you describe your version to be. If you do an analysis of a stereo signal, you will find that it can be rearranged into a sum and difference signal. The sum is mono, and the difference is the stereo information from both channels - ie, whatever isn't in the mono channel. Inversion does not enter into this, but subtraction does. The sum and difference equations are well-understood, and cannot be disputed, even by you.
And if you don't understand even this much, then you are doomed to fail, because there is an absolute prerequisite for being able to code anything - and that is that you understand the problem and issues that it raises in the first place. What amazes me is that you don't even seem to have studied how the CEP vocal cut works, and you are mouthing off about it as if you are the only person in the world who has ever thought of 'your' method. You are not. Not only that, but you don't seem to have anything like the right attitude towards investigating anything. |
I thought that I understood the vocal cut method perfectly - one of the two channels is inverted and mixed with the other, resulting in a mono track which, as you correctly say, contains the difference between the two channels. It can be done using the "vocal cut" preset in CEP, or it can be done manually. I never questioned the way it works, so why all this bashing?
If I'm totally and utterly wrong, would you be kind enough to enlighten us all about the way "vocal cut" DOES work?
Thanks for reading.
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Mon Jun 30, 2003 3:52 pm
|
|
|
| Graeme wrote: |
| Lendrik wrote: |
| With the centre channel extracted.... |
What centre channel? There is no centre channel - it's an illusion!
OK, I know we talk of it, but we are aware of the fact that it doesn't really exist and the term is conventinally used for convenience's sake.
There are two channels one left, one right.
That's it!!
No more than two,
no centre,
no way can you make it work.
|
Of course we know about that, Graeme! Do you think I'm an idiot? We are talking about the VIRTUAL centre channel, the one that gets eliminated when you do a vocal cut. I've said it sooo many times by now.
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Mon Jun 30, 2003 4:15 pm
|
|
|
| Lendrik wrote: |
| Do you think I'm an idiot? |
Out of the mouths of babes.....
Frankly - if you think you can actually do this - yes!
I will be the first to eat humble pie if you manage to achieve the desired end. In the meantime, I think you're off your trolley to think you can do something which anyone, who understands the theory, knows can not be done.
Hey, SteveG - do you think he's an idiot?
|
|
rfchmbrs
Location: USA
Posts: 30
|
|
Posted - Mon Jun 30, 2003 4:23 pm
|
|
|
I'm not thinking in any way of a centre/center channel. There is "signal" from all instruments on both recorded channels. The noise on both. Both signal and noise are time variant and change character in time and space. The matter comes to IDing the noise or the signal. I'll admit this is not simple to do but is being done elsewhere. The thing I find interesting in "music" processing is the lack of new ideas. Most all of what I see is an emulation of older analog hardware.
Let's not waste any of your time. I get paid to waste my time.
RON C
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Mon Jun 30, 2003 4:29 pm
|
|
|
| Graeme wrote: |
| Frankly - if you think you can actually do this - yes! |
I never said that I personally can do it, did I? If you are talking about the actual possibility of doing it, stating that it can't be done just because nobody did it so far is a bit shortsighted. If that's not the case, can you explain exactly why it can't be done?
|
|
rfchmbrs
Location: USA
Posts: 30
|
|
Posted - Mon Jun 30, 2003 4:34 pm
|
|
|
I forgot to mention that I'm a card carrying IDIOT and get paid for it. Venture capitalist have a lot of money for the idiots in the world.
I have questions of the vocal of you from above. Do you understand sound? Or do you just play with and listen to music? What do you know about ASW and other sound related technologies?
I will admit that with only two channels, the methods need more interaction.
Enough. Give it a rest.
RON C
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jun 30, 2003 4:35 pm
|
|
|
| Lendrik wrote: |
I thought that I understood the vocal cut method perfectly - one of the two channels is inverted and mixed with the other, resulting in a mono track which, as you correctly say, contains the difference between the two channels. It can be done using the "vocal cut" preset in CEP, or it can be done manually. I never questioned the way it works, so why all this bashing?
If I'm totally and utterly wrong, would you be kind enough to enlighten us all about the way "vocal cut" DOES work? |
Well, you keep doing it! If you invert and mix the stereo channels you don't get mono. You NEED bashing!
| Quote: |
| unlike the usual "vocal cut" track, which is a mono track with either left or right channel inverted and mixed together. |
It's not inverted, it's SUBTRACTED!
You want the math? Here it is:
Mono = Left + Right (M = L + R)
Difference = Left - Right (S = L - R)
Adding M and S signals gives:
M = L + R
S = L - R
M + S = 2L
Subtracting M and S signals gives:
M = L + R
S = L - R
M - S = 2R[/list:e4613d16ca] And by varying the magnitude of the difference signal, we can vary the width of the image. With the vocal cut, the onlyreason that we get an invert is to facilitate the final stage in the proceedings, which is the isolation. Check out two presets in the Channel Mixer. First, look at the LR to M-S preset and take a careful note of that - sign. Then do the sums, and work out what the vocal cut preset is doing. If you analyse this correctly, you will find that it does exactly what you are proposing. And it leaves everything that is dead centre, whateverit is.
Your problem is that you can't just remove the middle channel - because it contains information that makes the stereo signal work. If you remove it gradually, you will find that the remaining stereo image gets wider and wider until it doesn't make any sense at all. And this is why the only way to get the vocal cut preset to work is by reducing the final result to mono, which completely gets around the width problem.
I think that your problem is that you haven't realised the truenature of the illusion that is stereo. You can'tjust remove a chunk of it and assume that the rest will remain intact. And you don't have to take my word for this - you can try it for yourself. If you create a mono signal by adding two stereo channels together, and recreate this as two mono channels, you can then invert them from each side of the stereo signal. But you won't like the result much, I can assure you - and it will sound identicalto the effect of doing a vocal cut.
_________________
|
|
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jun 30, 2003 4:41 pm
|
|
|
| rfchmbrs wrote: |
| I have questions of the vocal of you from above. Do you understand sound? Or do you just play with and listen to music? What do you know about ASW and other sound related technologies? |
I don't go on about it, but I think that you'll find that I'm probably a damn sight more academically qualified, and experienced, specifically in acoustics, not physics, to comment on this than you are!
_________________
|
|
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jun 30, 2003 4:59 pm
|
|
|
| rfchmbrs wrote: |
| I'm not thinking in any way of a centre/center channel. There is "signal" from all instruments on both recorded channels. The noise on both. Both signal and noise are time variant and change character in time and space. The matter comes to IDing the noise or the signal. I'll admit this is not simple to do but is being done elsewhere. The thing I find interesting in "music" processing is the lack of new ideas. Most all of what I see is an emulation of older analog hardware. |
It's not being done any more successfully - at least not the results I've heard. The real problem is that humans use a psychoacoustic model to 'isolate' signals in noise, and that the paradigm that each individual uses is subtly different. Even if you could isolate the temporal/spatial signal for one individual, it wouldn't neccessarily work for another. Even though the ear's critical bandwidth follows a similar pattern in all humans, each individual varies - which is why some people get on better with MP3s than others.
I'll still say it's a waste of time, and money that could be spent on something more productive...
_________________
|
|
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Mon Jun 30, 2003 5:12 pm
|
|
|
| SteveG wrote: |
Well, you keep doing it! If you invert and mix the stereo channels you don't get mono. You NEED bashing!
| Quote: |
| unlike the usual "vocal cut" track, which is a mono track with either left or right channel inverted and mixed together. |
It's not inverted, it's SUBTRACTED!
|
What you said above is cotradicting with the post by Syntrillium M.D. (the most authoritative person on these forums) where he quotes our dear Graeme. Here it is: 'REMOVING VOCALS??'
Just a few quotes for you from that post:
| Quote: |
All these 'vocal cut applications work the same way. You can do it manually, using a wave editor such as CE, or you can rely on some other bit of kit which does all the steps automatically - but the end result is the same.
1 - Take a stereo file
2 - Invert the phase of one channel
3 - Add the two channels together
That's it - that's all you do. Now for what happens.
1 - Anything that was centre stage (hopefully, in this case, the vocal) will be cancelled out.
2 - This includes the bass, kick drum and anything else that was also centre stage.
3 - Material which is close, but not, centre stage will also reduce in level.
4 - The end file will be mono. |
|
|
ozpeter
Location: Australia
Posts: 3200
|
|
Posted - Mon Jun 30, 2003 5:33 pm
|
|
|
| Quote: |
| Syntrillium MD (the most authoritative person on these forums)..... |
Well, he's the person in authority, but not necessarily on all subjects the most authoritative - there's a big difference!
- Ozpeter
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jun 30, 2003 5:34 pm
|
|
|
| Lendrik wrote: |
What you said above is cotradicting with the post by Syntrillium M.D. (the most authoritative person on these forums) where he quites our dear Graeme. Here it is: 'REMOVING VOCALS??'
Just a few quotes for you from that post:
| Quote: |
All these 'vocal cut applications work the same way. You can do it manually, using a wave editor such as CE, or you can rely on some other bit of kit which does all the steps automatically - but the end result is the same.
1 - Take a stereo file
2 - Invert the phase of one channel
3 - Add the two channels together
That's it - that's all you do. Now for what happens.
1 - Anything that was centre stage (hopefully, in this case, the vocal) will be cancelled out.
2 - This includes the bass, kick drum and anything else that was also centre stage.
3 - Material which is close, but not, centre stage will also reduce in level.
4 - The end file will be mono. |
|
It's not a contradiction at all. If you work out what you've actually done to the files, you will effectively have inverted and mixed - which is a subtraction. I explained it in theoretically correct terms - but since you seem to have some trouble following/understanding that, I'm not too surprised that you couldn't work this out, either.
I am completely mystified as to how you could possibly think that you'd stumbled onto anything here - well, not about the 'stumbled', actually.
_________________
|
|
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Mon Jun 30, 2003 6:26 pm
|
|
|
| SteveG wrote: |
| well, not about the 'stumbled', actually. |
:D Nice one!
|
|
rfchmbrs
Location: USA
Posts: 30
|
|
Posted - Tue Jul 01, 2003 7:07 am
|
|
|
Steve,
I'm interested in your reasoning why separation can not be acheived so please e-mail me so we can take this off-line.
CHEERS,
RON C
|
|
bonnder
Posts: 215
|
|
Posted - Tue Jul 01, 2003 4:53 pm
|
|
|
Can this be done? Can someone give us a working model? Or at least suggest a time frame for when one will be available. This name calling sounds like it is coming from a bunch of people not very secure in their own skins. Will somebody please just create the product. And those who can't create the product ought to stop discussing it. As in, why insist that something is possible if you personally cannot make it happen.
----------
p.s. Please note - I am being slightly sarcastic & exasparated here.
|
|
MusicConductor
Location: USA
Posts: 1524
|
|
Posted - Tue Jul 01, 2003 11:08 pm
|
|
|
Until somebody can come up with software that can hear like an ear does, it's not going to happen IMHO. The software would need to discern the fundamental frequency of a voice separate from all other sounds, be able to recognize associated harmonics, and be able to "track" the constantly shifting frequencies involved. Plus, it would need to shift gears frequently to discern a multitude of non-harmonic sounds, such as vocal rasp and breath, as well as consonants. And if someone succeeded with this formidable task, then you'd still have residuals from reverb and other effects to contend with.
I'm glad it's not a job for me to be concerned with.
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Wed Jul 02, 2003 2:16 am
|
|
|
|
I know it well that isolating vocals from any other sounds in a recording is still a part of sci-fi. The subject of the topic you are reading, in fact, is a bit more trivial - trying to keep the sounds that are located in the middle of a stereo field (that usually includes vocals, bass, bass drum and snare drum) and getting rid of the sounds that were panned left or right. In other words, doing the opposite of what the "vocal cut" preset does. My first post in this topic gives a hint as to how exactly this may be achieved.
|
|
jonrose
Location: USA
Posts: 2901
|
|
Posted - Wed Jul 02, 2003 2:32 am
|
|
|
Right!
So... you could just use the Vocal Cut preset (for what it will provide) and be done with it, yes?
8)
man, this was getting truly painful!
_________________
 |
|
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Wed Jul 02, 2003 3:57 am
|
|
|
| Lendrik wrote: |
| The subject of the topic you are reading, in fact, is a bit more trivial - trying to keep the sounds that are located in the middle of a stereo field... |
Hohoho - Hahaha - tears are streaming from my eyes. I can't believe he thinks this is more trivial  .
Oh dearie me, ROTF - I think I'll go have a cup of tea to settle me down a bit .
|
|
William Rose
Location: USA
Posts: 467
|
|
Posted - Wed Jul 02, 2003 8:31 am
|
|
|
Speaking of The M.D., he hasn't written a post now for one day shy of a fortnight.
Because, it occurred to me, if he was here he probably would have shot this thread behind the ear by now.
I don't know though. I know he has a life apart from this forum but, is it typical for him to vanish for weeks at a time ? Course, vacations (holidays) sometimes last two-weeks.... Or so I've heard.
And just where does somebody who lives in Arizona go for a summer holiday ? Somewhere cold ?
|
|
bonnder
Posts: 215
|
|
Posted - Wed Jul 02, 2003 8:37 am
|
|
|
Seems to me in the relatively recent past MD made reference to packing things up because he was moving. Reference was maybe in the "What Do We Look Like" thread (title may not be dead on accurate)? I took that to mean that staff shifting was already occuring at Syntrillium in Arizona.
---------
Nope - it's at "Pictures of Your Studio" :
'Pictures of your studio'
|
|
MusicConductor
Location: USA
Posts: 1524
|
|
Posted - Wed Jul 02, 2003 9:34 am
|
|
|
Yep, I miss the M.D., too. Support has been doing a great job keeping up with the essential help, but the Doc's good banter adds to the warmth of this cyberplace.
Back to the isolating of the "center" information, which in this part of the world is called a "phantom center channel" because of the illusion that the sound appears between your speakers when everything's done right. So, how many patents does Dolby have? If you want to check out what's already been done in this field, study up at www.dolby.com and elsewhere just how they derive the center channel in a Pro Logic matrix from the two input channels (called Lt - Rt). Phase, steering, etc -- and guess what? It's not perfect, and a little of the center information and all the stereo reverb and effects remain in the other channels. But it might be worth your looking into that, Lendrik.
|
|
2Bdecided
Posts: 340
|
|
Posted - Thu Jul 03, 2003 5:47 am
|
|
|
Lendrik,
You can't perform binary operations, in the way you suggest, on audio data. It doesn't work. If you doubt this, please try it. What you're suggesting is trivial to program - try it and see what comes out!
Cheers,
David.
EDIT: can't spell.
|
|
rfchmbrs
Location: USA
Posts: 30
|
|
Posted - Thu Jul 03, 2003 7:55 am
|
|
|
Re: Dolby.
What Dolby does has to be real time and fully automatic (coded in a DSP chip). What can be done on a PC/MAC does not have to be real time and can also use human interaction/training. After enough training an algorithm might fall out that is "automatic". Not matter what algorithms we are talking about, the separation is probably less that 30 of 40 dB. Creating (virtual channels) and removing (vocal other noise) information are different algorithmically but are close cousins.
RON C
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Thu Jul 03, 2003 3:13 pm
|
|
|
I am very thankfull to the last three posters for providing intelligent replies that at least make sence, and not laughing me off like some do. Obviously, I'm not a professional sound engineer and don't have the knowledge of things that they consider trivial, so forgive me if my posts look stupid.
To reply to 2Bdecided, I'd love to perform all kinds of experiments on audio, but I don't have enough knowledge of programming to make my own program. If somebody could supply me with a simple piece of software that could perform functions on a wave sample by sample, I'd even pay for it.
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Thu Jul 03, 2003 4:18 pm
|
|
|
| Lendrik wrote: |
| I am very thankfull to the last three posters for providing intelligent replies that at least make sence, and not laughing me off like some do. |
You've had plenty ofreplies that make sense - you simply ingnored them.
| Lendrik wrote: |
| Obviously, I'm not a professional sound engineer and don't have the knowledge of things that they consider trivial, so forgive me if my posts look stupid. |
Many of the replies you have had have been from professionals - guys who actually do have the knowledge which you admit to not having. However, that hasn't stopped you from continuing to perpetrate your stupid idea.
You originally said "Well, it may not be possible on Cool Edit (at least at the moment) but it is possible in theory." More than one person has told you that - even in theory (if you understand the theory, that is) it is not possible.
Why do you persist?
|
|
ozpeter
Location: Australia
Posts: 3200
|
|
Posted - Thu Jul 03, 2003 4:23 pm
|
|
|
Within the last year, someone posted here the results of hours of work they'd performed on a small part of a carefully selected track, using every conceivable tool (including many carefully-tuned passes of noise reduction), in an attempt to isolate the vocal. The result was, as a demonstration of ingenuity and determination in the face of overwhelming odds, impressive. In audio terms, however, it sounded like an indifferent recording of a strangled robot.
- Ozpeter
|
|
AMSG
Location: Sweden
Posts: 594
|
|
Posted - Thu Jul 03, 2003 4:59 pm
|
|
|
|
Hehe, is this thread still going?
|
|
ozpeter
Location: Australia
Posts: 3200
|
|
Posted - Thu Jul 03, 2003 5:07 pm
|
|
|
Well, AMSG, the theory was that keeping two or three threads on this subject running at the top of the forum might deter any more new ones. Sadly, the plan failed..... pity, for so many months we'd done so well in avoiding them. Still, every horse loves a few jumps over some familiar fences every now and then. ("Not me!" I hear Graeme saying...)
- Ozpeter
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
jonrose
Location: USA
Posts: 2901
|
|
Posted - Thu Jul 03, 2003 7:59 pm
|
|
|
Actually, I think you might actually have doused the flames, somewhat.
We already know that either eliminating or extracting one thing from a mix can't be perfectly done... what you've posted is just a reference to a method of extracting the "center" information with a different program (SF). You could certainly perform this kind of function in CE, too - the steps are somewhat different, is all, and they've been detailed on this forum in the past.
(Now, my post probably will throw wood on the fire - I can just see some newbies coming along and construing it as condoning vocal cut/extraction... )
ROFL!!!
Best... -Jon
_________________
|
|
|
|
VoodooRadio
Location: USA
Posts: 3971
|
|
Posted - Thu Jul 03, 2003 8:18 pm
|
|
|
| Quote: |
| Sadly, the plan failed..... pity, for so many months we'd done so well in avoiding them. |
Patience.... my young Jedi. The start of the school year is just around the corner and things will settle back down once the kiddies get back in class. 
_________________
I said Good Day!
Voodoo
 |
|
|
|
kevino
Posts: 10
|
|
Posted - Thu Jul 03, 2003 8:42 pm
|
|
|
I know that this forum is littered with posts that explain this topic, and how stereo is created and the phatom center channel and all that stuff, but is there some website that explains all this in plain english? I really want to understand all this, but though I completly believe everything that SteveG & co are saying, i sadly don't understand most of what they say on the forum (esp L-R, etc.) I would greatly appreciate guidance to a good site (or book, even ) thank you
KO
|
|
MusicConductor
Location: USA
Posts: 1524
|
|
Posted - Thu Jul 03, 2003 10:49 pm
|
|
|
I was hoping Syntrillium's books page would be of help, but I really think something simpler is in order. Do have a look at the glossary in the Help file, the online tutorials, and keep lurking here!
| Lendrik wrote: |
| I am very thankfull to the last three posters for providing intelligent replies that at least make sence, and not laughing me off like some do. |
Um, well, the whole time I was writing about that Dolby thing, um, I was laughing.
:D
Seriously, it is a technology that may have some merit for you. However, it is also another illustration from a different approach that shows just how weak these processes are. Dolby has made millions on a technology that is just effective enough to be of some value, yet is amazingly limited in its abilities. And so it will ever be.
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Fri Jul 04, 2003 12:56 am
|
|
|
| kevino wrote: |
| I know that this forum is littered with posts that explain this topic, and how stereo is created and the phatom center channel and all that stuff, but is there some website that explains all this in plain english? I really want to understand all this, but though I completly believe everything that SteveG & co are saying, i sadly don't understand most of what they say on the forum (esp L-R, etc.) I would greatly appreciate guidance to a good site (or book, even ) thank you |
That's a fair point, but there is going to be a little difficulty here - because in order to fully understand what is happening, you need to have a grasp of some of the basic concepts of sound propagation and the construction of wavefronts. Without at least an intuitive grasp of these fundamentals, it is going to be very hard to synthesise an understanding of a second-level concept. And unfortunately, this subject falls very firmly into the second-level category.
I could try several different forms of words - and some of them may have some resonance with you, I suppose... but to explain the whole subject in plain english would take up a lot of space, and would probably be counter-productive anyway. The real problem is that an explanation of this subject relies on concept formation, and without being able to question you on precisely what you do and don't understand, it wouldn't make sense to move on to the next part.
I don't know about anybody else, but I haven't ever seen an extended explanation of the issues behind this in plain english - I think probably because the majority of people who write about it take a polarised stance. They either understand the acoustic principles involved, and explain it in these terms, or think that they've found the answer - which they invariably haven't - and write gobbledygook, often about tracking filters, boolean logic, throwing large amounts of DSP at it, etc etc.
So my answer to you really is that it's probably not this particular concept that you need to understand initially - it's a few others.
_________________
|
|
|
|
AMSG
Location: Sweden
Posts: 594
|
|
Posted - Fri Jul 04, 2003 3:46 am
|
|
|
| ozpeter wrote: |
Well, AMSG, the theory was that keeping two or three threads on this subject running at the top of the forum might deter any more new ones. Sadly, the plan failed..... pity, for so many months we'd done so well in avoiding them. Still, every horse loves a few jumps over some familiar fences every now and then. ("Not me!" I hear Graeme saying...)
- Ozpeter |
Yes, the plan DID fail. But you're right. It was very quiet a while...
Hehe, and steve and graeme seem to be near a nervous breakdown, the poor guys! Maybe we should come up with a new interesting, challenging thread for them, so they can answer to that one? That way they can start to repress the bad memories of this one?
|
|
ozpeter
Location: Australia
Posts: 3200
|
|
Posted - Fri Jul 04, 2003 4:49 am
|
|
|
http://www.ethanwiner.com/novocals.html was my best shot in the 'sticky thread' (as I write this, just about to hit 7000 reads!) at providing a newbie-friendly explanation.
- Ozpeter
|
|
2Bdecided
Posts: 340
|
|
Posted - Fri Jul 04, 2003 8:24 am
|
|
|
ozpeter,
It's a very good explanation (I'm not just saying that), but I have two criticisms.
Firstly, what's the point of 7 steps? you can do it in CE in two:
1. apply low-end shelf cut to the left channel only
2. Use channel mixer vocal cut
That's exactly the same, isn't it? If you want mono, now convert to mono.
Secondly, the "mix to mono then invert mix paste back into the stereo version" is exactly the same as using the vocal cut pre-set - but with one channel inverted. That's all. No stereo information remains.
Lendrik,
so you don't know anything about audio, and you don't know anything about programming - but you're suggesting that someone should program this audio processing idea you've had. An idea that everyone else just happens to think is impossible? There's a name for you - optimist!
;)
Cheers,
David.
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Fri Jul 04, 2003 4:04 pm
|
|
|
| 2Bdecided wrote: |
| Lendrik, so you don't know anything about audio, and you don't know anything about programming - but you're suggesting that someone should program this audio processing idea you've had. An idea that everyone else just happens to think is impossible? There's a name for you - optimist. |
Actually, I know quite a lot about audio and sound processing for an ordinary guy like me. I studied Access To Music Technonlogy, that's a course you take prior to going to University, and also a weekend course in sound engeneering. As for programming, I can program in QBasic and a little bit in Visual Basic, but I don't know how to make a wave editor.
I've got a more or less clear idea of what can be done to exctract the middle-channel sound from a stereo, from a programming point of view; all I need is the means to do that - a software that would read from a stereo wav file, compare the sample values of the left and right channels to each other and create an output wave. You may understand why I will not reveal everything about my method here on these forums, but I've got a pretty good idea of what I want to do.
|
|
post78
Location: USA
Posts: 2887
|
|
Posted - Fri Jul 04, 2003 4:08 pm
|
|
|
| Quote: |
| You may understand why I will not reveal everything about my method here on these forums, but I've got a pretty good idea of what I want to do. |
It makes no difference; your method is wrong. But hey, whatever it takes to get you to realize this faster...
_________________
Answer = 1. Probably.
|
|
|
|
Graeme
Member
Location: Spain
Posts: 4663
|
|
Posted - Fri Jul 04, 2003 4:12 pm
|
|
|
| Lendrik wrote: |
| I know quite a lot about audio and sound processing for an ordinary guy like me. |
However, you obviously don't as much as you think you do - else you wouldn't be pursuing this impossible 'holy grail'.
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Fri Jul 04, 2003 4:41 pm
|
|
|
| Lendrik wrote: |
I've got a more or less clear idea of what can be done to exctract the middle-channel sound from a stereo, from a programming point of view; |
Yeah, I'd sort-of agree with that. Well, the 'less' bit, at any rate... The rest of what you said is supposed to be some kind of 'joke', I take it...
_________________
|
|
|
|
Lendrik
Location: United Kingdom
Posts: 17
|
|
Posted - Fri Jul 04, 2003 4:45 pm
|
|
|
| Graeme wrote: |
| However, you obviously don't as much as you think you do - else you wouldn't be pursuing this impossible 'holy grail'. |
Maybe, but that's exactly how great discoveries are made. 
My only "vice" is that I fail to see why this goal is impossible to achieve. I will keep on trying until I myself will explicitely and exclusevely see why. I perfectly accept that my idea might prove to be totally wrong, but out of curiousity I would like to see what the real result of my method will be.
|
|
VoodooRadio
Location: USA
Posts: 3971
|
|
Posted - Fri Jul 04, 2003 4:53 pm
|
|
|
Get back with us after you've got it all figured out. Oh, and..... have fun!!!
_________________
I said Good Day!
Voodoo
|
|
|
|
2Bdecided
Posts: 340
|
|
Posted - Mon Jul 07, 2003 5:04 am
|
|
|
| Graeme wrote: |
| Lendrik wrote: |
| I know quite a lot about audio and sound processing for an ordinary guy like me. |
However, you obviously don't as much as you think you do - else you wouldn't be pursuing this impossible 'holy grail'.
|
Whilst I think Lendrik's idea is probably going to fail, I don't think you should say that anything is "impossible".
We do the impossible every day with Cool Edit. Well, impossible by the standards of 50 or 100 years ago. If you go back that far, the theory and mathematics was unheard of for most of the filtering etc we carry out, let alone the computing power!
As for vocal removing/isolating: I'd agree that it might not be perfected within my life-time, but there are already "auditory attention processing" simulations which aim to break-down auditory sound scapes in the same way that the human ear/brain do. Once you've figured out how auditory streaming works, you've designed a vocal remover tool without even meaning to.
There was a great PhD on-line about Auditory Scene Analysis, with sound samples of an algorithm's attemps to perform the task. There was a recording of a busy street, and the algorithm separated the traffic rumble from car horns from people talking etc. I can't find it anymore, but it was quite impressive - no where near perfect, but things have to start somewhere.
Cheers,
David.
|
|
|
|
2Bdecided
Posts: 340
|
|
Mark T
Location: Norway
Posts: 890
|
|
Posted - Mon Jul 07, 2003 5:47 am
|
|
|
| SteveG wrote: |
| rfchmbrs wrote: |
| I'm not thinking in any way of a centre/center channel. There is "signal" from all instruments on both recorded channels. The noise on both. Both signal and noise are time variant and change character in time and space. The matter comes to IDing the noise or the signal. I'll admit this is not simple to do but is being done elsewhere. The thing I find interesting in "music" processing is the lack of new ideas. Most all of what I see is an emulation of older analog hardware. |
It's not being done any more successfully - at least not the results I've heard. The real problem is that humans use a psychoacoustic model to 'isolate' signals in noise, and that the paradigm that each individual uses is subtly different. Even if you could isolate the temporal/spatial signal for one individual, it wouldn't neccessarily work for another. Even though the ear's critical bandwidth follows a similar pattern in all humans, each individual varies - which is why some people get on better with MP3s than others.
I'll still say it's a waste of time, and money that could be spent on something more productive...
|
I have only read the abstract of the paper. As far as I understand it the attempt here is to duplicate (or at least understand) the way in which people "sort" out a jumble of sounds in their minds, NOT to separate them into high quality tracks. I think people get confused by terms like "isolate", which makes them believe that the mind can physically separate the sounds. All we can do is focus on, or pay attention to, individual sounds which we recognise and analyse through various cognitive mechanisms.
_________________
Mark 
nil desperandum - nunc est bibendum |
|
|
|
SteveG
Location: United Kingdom
Posts: 6695
|
|
Posted - Mon Jul 07, 2003 6:10 am
|
|
|
I'm aware of that research. A couple of points - consider the implications of section 2.3.2, and then ask yourself why there are no musical examples... the correlation issue is huge, and pretty much insurmountable. Despite extensive research, I'm not aware of any significant breakthroughs at all. Any updates from anywhere?
With a little care, CEP's noise reduction can do almost as well as his PhD model! But it's certainly a good piece of research, precisely because it explores the limitations of the approach, and treats them realistically.
_________________
|
|
|
|
2Bdecided
Posts: 340
|
|
Posted - Tue Jul 08, 2003 9:04 am
|
|
|
| SteveG wrote: |
| I'm aware of that research. A couple of points - consider the implications of section 2.3.2, and then ask yourself why there are no musical examples... the correlation issue is huge, and pretty much insurmountable. |
That's probably true - but in the current context, it may not matter - let's say we can isolate the vocal quite well, but there's a few extra harmonics added by some instruments at some moments. I can hear that it's wrong. You can hear that it's wrong. Therefore, a computer model of the human voice can also hear that it's wrong, and fix it.
| Quote: |
| Despite extensive research, I'm not aware of any significant breakthroughs at all. Any updates from anywhere? |
Not seen any.
| Quote: |
| With a little care, CEP's noise reduction can do almost as well as his PhD model! |
That's very true. I've used it in similar situations where much-less-than-perfect separation was good enough. It's a painfully slow approach.
| Quote: |
| But it's certainly a good piece of research, precisely because it explores the limitations of the approach, and treats them realistically. |
I'll agree that complete isolation is probably impossible. OK, must be impossible. What about "good enough that a normal person doesn't perceive any faults?" We're not talking about separating a violin from a cello (something that many normal people can't do with their ears anyway) - we're talking about separating out the human voice from other very different sounds.
Maybe the results will always sound like a strangled robot. But I don't see any theoretical grounds to say that this must be the case. Maybe I'm an optimist too ;)
Cheers,
David.
|
|
rfchmbrs
Location: USA
Posts: 30
|
|
Posted - Tue Jul 08, 2003 9:18 am
|
|
|
What is an acceptable suppression? 10, 20, 30, 40dB?
Being an optimist, I'd say it is possible to get at the most 30dB.
Anyone here ever work with High Ordered Correlation methods? Bi-Correlation, Bi-Coherence etc. Also the complex spectrogram?
RonC The Physicist
|
|
| |
Topic
|
All Forums
My idea of isolating vocals (centre channel)
