Talking about Mixing Revolution: freq collision vs "masking"

[pumafred]

This may be too off topic for this forum, but I am curious about the difference between "frequency collision", which can be detected, e.g., with MultiAnalizer, and "masking", which I only became aware is a different thing on some discussions about Neutron. (Izotope claim they are dealing with "masking" in Neutron.)

Today I came across this, which raises the point. I am sure there are people smarter than me in this forum who may be able to think of some way to use this info with Melda products to facilitate mixing:

http://science-of-sound.net/2016/09/aud ... dont-hear/

[MeldaProduction]

Well, first of all, I strongly doubt Neutron does anything about masking. For all I know the only "new" thing was that it shows an analyzer for 2 tracks and lets you adjust eq for both. Not exactly groundbreaking thing if you ask me.

Anyways currently the basic masking (as described in the article) and collisions are quite similar. We simply assume that if you have 2 instruments playing at the same frequency, one of them just will win and the other will be sort of covered by dirt. The masking as defined in the article also states that a frequency doesn't collide only with the same frequency, but also surrounding mainly higher frequencies. People usually deal with it with an eq, which i static, bad imho. MSpectralDynamics can do that dynamically and with smoothness it can actually cover surroundings as well. MAutoDynamicEq can do that as well, but you need to find the problematic frequencies manually. (does Neutron have a dynamic eq like this? dunno)
In most cases people deal with this only if they actually hear a difference, which is, well, correct imho, but rather hard to do.

For the mentioned "revolution" I actually plan a fully adaptive automatic solution, that's all I'm going to say now .

Anyways the eternal problem is that the brain doesn't react only to frequencies in similar region, but also spectral patterns - like if you hear a powerful sound that takes all your attention. And that's something impossible to solve imho, since it is supercomplex, would require AI of some kind, and mainly it is different from person to person.

[pumafred]

Heh heh, I knew I shouldn´t have mentioned Neutron...

Anyway, they say they are doing something about masking, and even point out to a paper they presented about this at the Audio Engineering Society.
Unfortunately, the article is not free for non-members, so this is the only thing I have:
http://www.aes.org/e-lib/browse.cfm?elib=18450

To clarify, I was not interested in discussing Izotope, only masking. Thank you for your quick and clear response.

[MeldaProduction]

Hehe yeah teasing me, right?

Anyways my pleasure .

[pumafred]

I think you have to move with the Mixing Revolution, since, from what I see on the web, the "new thing" is different forms of resonance detection, automation, artificial intelligence, etc. (For what they may be worth, Oeksonic, Sonible, TDLabs, Izxxxxx , etc.)
Many who are more interested in making music than in twisting knobs, and "sound creation", will go for those products. And I think you have everything to be very ahead in this area. The Turbos are all fine, but I think "automated mixing", or "AI assisted mixing", is the future.

However, perhaps it is only me who is interested. I've always been fascinated by the possibilities of AI, without thinking it could do everything.

[spencerlee]

Well why not go the whole way and have automated music creation, lyric creation, automated marketing, money printing and computer generated food. We should really get rid of human interaction in the Arts all together and let the computers do it as they are much more artistic and will make much better decisions.
Spencer

[pumafred]

Oh, so tiresome. Afraid of vacuum cleaners, washing machines, automatic transmission in cars, GPS, equalizers, compressors, digital audio? Do not fear. Human creativity will always be essential. But some people will think many routine tasks hamper their creativity, and it would be interesting to leave to computers.
Still, nobody will stop you from working with old audio technology, if it makes you feel more creative!
F
PS: If I wanted to be provocative, I would say that automated music creation (Brian Eno?), automated lyric creation, and all the others you mention, are fascinating. Human curators will always be needed to make them better. But, of course, I am only 60 years old. You must be older, I guess.

[Chandlerhimself]

I always thought of masking as completely or almost completely covering a sound while collisions are just sound covering the same frequency. I don't think masking can be reliabley detected. Just because 2 things are colliding doesn't mean they are masking each other. For example a synth pad and a snare drum usually don't mask each other despite the fact that they cover the same frequencies.

I doubt there will ever be anything that can mix for you, but I hope the mixing revolution can speed up some of the more tedious processes and let me get to the more creative aspects faster.

[pumafred]

I doubt there will ever be anything that can mix for you, but I hope the mixing revolution can speed up some of the more tedious processes and let me get to the more creative aspects faster.

I always think of mixing like cooking. Very few robots will consistently beat great chefs, but many robots will probably cook better food than most of us. (And then we can add a bit of salt, pepper, and other, to our own taste.)
Anyway, complex topic.

[spencerlee]

I absolutely love where technology is now but using technology to make creative decisions will end in tears.
If you cant hear frequency clashes you can see them on a spectrum analyser.
Put an analyser across your master bus and you can check instantly. Sidechaining is your friend.
The decision made on eq is a creative one not something to be left to a computer.
Orchestral arranging relies on different instruments on similar frequencies.

There are tasks that can be delegated but not most of the mixing ones. I have made presets for every plugin I create but they all have to be tweaked to make them fit the particular piece of music I am working on.
A limiter at the end of the chain is something that works automatically and is necessary.

I find things like getting the phase coherency correct on drum mikes by moving the phase destroys the sound of the drum kit. Technology gone mad. It is either in or out of phase.
Autotune is technology gone mad but I still use it. Telling singers they do not need to learn to sing in tune is degrading the skillset.

I cant actually see how the mixing revolution can possibly replace any of the great mixers. Therefore anything the mixing revolution does that you don't have to learn is stopping you from becoming a great mixer.
I see automated mixing as "Striving for Mediocrity"
Spencer

[jmg8]

Autotune is technology gone mad but I still use it. Telling singers they do not need to learn to sing in tune is degrading the skillset.

I cant actually see how the mixing revolution can possibly replace any of the great mixers.

I didn't hear of any singers that stopped practicing because autotune came along.

Don't think that these new tools are coming to REPLACE you but rather AUGMENT you.

[MeldaProduction]

pumafred: I'm actually not really afraid. The audio industry imho isn't moving very fast, it at all . What I have planned is such a big step, that I'm more like afraid if people would be ready for it . But I have plans for that too. The MTurbo things are actually sort of the prerequisite for all of it, I'm gathering technology for a long time now . I think we have nearly everything now, and now the final stage of the masterplan (which I really hope won't end up in s***, since there are still quite a few challenges...)

Btw. as for the AI - for some reason this really seems to be "in" now and everyone is using it for marketing, yet I have never seen any product, that would actually do anything even remotely close to anything, which would be remotely close to anything, which would be remotely close to an AI . For instance Russians are now hyping about the "AI military robots", which are based on deep neural networks... A little secret here - deep neural networks is the keyword, marketed as AI these days, but it is anything BUT it... It is just a nice name for a complex approximation method based on observations of neurons in our brain, but it misses one pretty important thing: it doesn't learn! You basically need to teach it ahead, which takes a lot of power and is severely limited, and then use it for actual calculations, but at that moment it doesn't learn anymore. As far as I know nobody made the simultaneous application & learning work well yet. I saw some google stuff recently, maybe they made some progress, but it didn't exactly look like a breakthrough yet .

Spencer: Who said anything about replacing humans? I simply want to move the technology forward, so that the bold people can finally trash the analog stone-age and try something that would hopefully make it much easier, faster and better sounding. Needless to say, there will be an operator, but hopefully he will have much less work and won't need to do the same thing all the time . I rather focus on the music than on the "work" around it.

[alexis1]

Well, first of all, I strongly doubt Neutron does anything about masking. For all I know the only "new" thing was that it shows an analyzer for 2 tracks and lets you adjust eq for both. Not exactly groundbreaking thing if you ask me.

Anyways currently the basic masking (as described in the article) and collisions are quite similar. We simply assume that if you have 2 instruments playing at the same frequency, one of them just will win and the other will be sort of covered by dirt. The masking as defined in the article also states that a frequency doesn't collide only with the same frequency, but also surrounding mainly higher frequencies. People usually deal with it with an eq, which i static, bad imho. MSpectralDynamics can do that dynamically and with smoothness it can actually cover surroundings as well. MAutoDynamicEq can do that as well, but you need to find the problematic frequencies manually. (does Neutron have a dynamic eq like this? dunno)
In most cases people deal with this only if they actually hear a difference, which is, well, correct imho, but rather hard to do.

For the mentioned "revolution" I actually plan a fully adaptive automatic solution, that's all I'm going to say now .

Anyways the eternal problem is that the brain doesn't react only to frequencies in similar region, but also spectral patterns - like if you hear a powerful sound that takes all your attention. And that's something impossible to solve imho, since it is supercomplex, would require AI of some kind, and mainly it is different from person to person.

Great, informational, useful response and summary of masking vs. collision.

Slightly off topic, if I had a parallel life, I would try to do a rigorous quantitative analysis of how collision -and- masking differ between vocal harmonies generally considered to be excellent, vs. poor quality, perhaps amateur, vocal harmonies. It seems it is generally accepted (and I concur with) that in a two-voice harmony model, it usually sounds best if the two voices are qualitatively quite different to each other (one sandpaper rough, one sweet and smooth, like Everly Bros., e.g.). I'd love to see how some as yet to be determined masking/collision parameter would differ from the same passage being sung by one person overdubbing!

[spencerlee]

Hey Vojtech I agree with trashing the old analog history. As I have said before I came through the early 60s when analog was supposed to be great. We all thought the equipment was not up to scratch. Always worried about noise and distortions.
A keypex gate on every tape channel to get rid of noise or the degradation of Dolby or DBX noise reduction.
Most of the old records sound like crap now.

What the electronic designers got right was EQs, Compressors and Microphones. The electronic components quality did not live up to the designs mostly. I have built a Microphone that rivals any of the vintage mikes. It has no transformer but uses a tube and has a preamplifier built into the power supply using state of the art intergrated circuits (ic's). The capsule is from 797Audio in China.
I put my mike up alongside my U67 and I prefer my mike. It is not how the mike sounds on its own, it is what happens to the sound when you put it in the mix.

The design of MTurboEq and MTurboComp certainly are an improvement on the original hardware they simulate. For a start no transformers. Check youtube and you will see most of the major engineers are now mixing in the box with the old hardware there as show to impress clients. It makes me sad when I see people paying ridiculous prices for vintage crap.
I am very up for anything that will make my workflow easier and is the reason I turn any chain I use all the time into a plugin in MTurboMix. Vojtech you should release MTurboMix because it is just wonderful.
Spencer

[vectorwarrior]

Just wanted to +1 the talk about an exciting future that goes beyond the obsession with vintage gear. We should strive for more than reliving the past. I'm very excited about the mixing revolution, in whatever form it takes!