So far I'm not seeing tons of videos on depth into using Spectralayers 11. Obviously 8 and 10 are similar, but I really think there is a big audience for people wanting to go very deep into using this tool. I think that there are plenty of video production uses for stems, and many stems also involve separating out some individual instrumental line, or as you showed specific speech lines out of say crowd noise, or outdoor noise. So, I'd like to learn how to find, and move tiny bits of spectral material out of one layer and into another, how to create custom layers (I assume that's a thing). I'm liking the way you do these videos sort of diving in, and playing around, while editing up a decent documenting video of your adventures. I sort of assume you do a wide range of stuff there in Warsaw (I think I heard you are from Warsaw).
I go all the way back to version 7 pro. I use it along with izotope RX 11 advanced and lots of plugins for forensics. For heavy duty buried voices- Hand carving using a spectral scope and manual brush is still the only answer, specially when the voice is only a distant part of the signal. Only issues so far is the GPU (4090) optimization is useless. Supposed to be fixed in a patch…
very good video regarding ( add this as plugins in Nuendo ) I think if they take a new path will be even better like enabling direct offline processing, and add these modules inside proceses will be something huge as most of these modules are not made to run in real time and they can also improve the experience of direct offline processing to make it easier to work with
I've actually liked better the SL dereverb than Acon's. The voice sounded fuller and I think a second pass with Acon or Clear (also great for denoise/dereverb) it actually may get almost into studio quality territory. Great video, very entertaining.
Hi there, can you explain your request for destructive editing please? Surely the result would be the same, because that’s the best separation it can do, with the only difference being that whatever it’s decided becomes permanent. With non-destructive, it makes its decision but you have the option to move things manually from one layer to another because everything’s still there. Or am I missing something? Would the destructive method allow for a more powerful algorithm?
That was my thinking too. Destructive vs non-destructive would simply have less storage use, but the effect of competent separation really has to do with underlying algorithmic assignment of the many tens of thousands of spectral "pieces" to their role ("bass", "drums", "vocal", "synth", etc), all of which is done using a set of computer "tricks" or algos (a bit like doing the square root, or doing a sort, only this is likely far more complex and is operated upon the underlying detailed sound analysis, which, oddly, I understand comes out of naval SONAR tech). I did some programming, and we developed algos to handling data the way we had to process it (mine was financial data). Here we know the start is a controlled breakdown using Spectral Analysis (there are menus choosing the exact settings for this) or to call it by its Comp Sci term a window of Fast Fourier Transform analysis (FFT) which then is uniquely identified piece by piece and generates this enormous number of fragments. A big part of assignment would be same frequency, but that is not the be-all, end-all. A singer may be singing a held note of the same frequency as a flute playing along. What probably counts more is start of note and then related hold of it and release, and identifying transients and formants by type. Then assigning that a tentative assignment as a particular instrument or vocal. In this case the components of the sound are like strings of items numbering in large numbers and getting assigned to categories. Once assigned we have these "lanes" we can listen to.
Ran unmix crowd, music, instruments including drum kit elements (Kick, Snare, Cymbals and Toms) on a 40 year old bootleg. Not perfect but very, very good. The next step will be to try the 'transfer' function where you find, say, a lead vocal that is in the 'Crowd' stem and graphically lasso it and transfer it to the Vocals track. No so quick, but should sharpen the stems up enough to basically remix a dirty, tape 'master' recording into something more balanced and listenable. A great leap for Spectralayers, and the next version should see them out of sight of the competition. Cool Vid Paul, thank you.
This is the next thing I want to learn too. I have stems with Ella Fitzgerald's singing and her back up band (on the "Songbook" albums, say the tune "Yesterdays"), but a held note from Ella was so steady the algo seemed to think this was a flute, and put it into the other layer. So, getting Ella's held note back in the vocal would be nice.
@@FoliaSound As a sometime programmer, I'd think destructive vs non-destructive is not really going to guarantee far better inclusion/exclusion. Instead I think what this entire product design is winds up being a mass of FFT analysis is done on the source, then it creates a giant array or similar store of "pieces" of spectral info, which is categorized as to which layer. It is this categorizing algorithm which we musicians sense. We know what content goes into a bass part, or a vocal. The algo would have to use what we might call "tricks" to assign fragments into categories. An example of an algo or trick would be formant type human voice sounds - our speaking and singing starts and ends often with formant type items from our mouth parts, so the pieces with those are going to get assigned to vocal (or speech, if that choice is made), then the algo would look at the sequential material right next in time and at the same frequency as this obvious "formant" stuff, and assign that to "vocal", and so forth. That's a highly simplified way of describing the algorithm, but it has to be something akin to that. A set of rules used to assign "type" to the broken down spectral content which is now separated into thousands of pieces. This would be the same regardless of if it is all stored separately or if the original file is wiped out and separate track files are made. Today we have so much storage around that it's not a big deal to just hang on to vast amounts of underlying data in most cases. There are exceptions, but it is true there are now often Terabytes of places to store data.
We are now in our post pro studios completely on Spectralayers. Way better results and great workflow for audiorestauration too. RIP izotope RX. SL 11 highly recommend!
I agree, at least with denoising. A year ago, almost all my denoising went through RX, but now Clear, Acon, dxRevive and SL took the mantle. The only aspect RX still is king to me is spectral editing and vertical cleaning, and some niche tools like de-wind. Exciting times for voice processing!
Not sure how you got unmixing of multiple voices to work but I've no success with it at all. It's the reason I upgraded to this version and it does not work on any songs I've tried (after separating out the vocals and chorus first of course). Nice video though.
Well, you took perfectly recorded voice recording and mixed it with noise and stuff…. Thats not how it goes, and those tests are cool, but not that realistic…. Still, spectralayers is A+
so nonsens....you take take 2 voices , let them see in a mix and then unmixed.......... what did happend in between , how did you unmix them , what was the proces?
i would definitely NOT recommend buying this. the unmix module can't even separate out the most basic stems. if you try to unmix a basic pop song, spectra won't be able to discern between a kick drum and a snare drum. it's just a giant unusable disaster - same as all unmix algorithms before it. nothing groundbreaking here unfortunately, and now i spent 300 bucks on this useless software because there was no trial... feels like that was probably by design. if you're looking for a way to separate stems from a song, this aint it.
It's a BULLSHIT toy for all those todays wanna-be "producer". Real producers don't need something like that. Or do you see Snoop, Dr. Dre, Quincy Jones, Rick Rubin or ANY real producer using that shit? Of course not.
Hey, have you heard anything about a thing called audio post production for multimedia? Have you ever thought, that working in a DAW is not only about producing music, but much more?
Helllooo! What do you think about new Spectraalayers 11? Do you have something else as your spectral suite? Lemme know down below, thanks!
What do you think about the declip and declick Tool in spectralayers compared to RX 11. Is it worth switching now for audio Post Produktion?
Big fan of the idea of destructive/non destructive options.
Remember you can always go into the other layers and transfer the sound you’re missing or wanting to get rid of.
can you elaborate some more on that please
@@rolisaenz it may be easier to look up “spectralayers cut to layer above”
@@rolisaenz yes, please elaborate
So far I'm not seeing tons of videos on depth into using Spectralayers 11. Obviously 8 and 10 are similar, but I really think there is a big audience for people wanting to go very deep into using this tool. I think that there are plenty of video production uses for stems, and many stems also involve separating out some individual instrumental line, or as you showed specific speech lines out of say crowd noise, or outdoor noise. So, I'd like to learn how to find, and move tiny bits of spectral material out of one layer and into another, how to create custom layers (I assume that's a thing). I'm liking the way you do these videos sort of diving in, and playing around, while editing up a decent documenting video of your adventures. I sort of assume you do a wide range of stuff there in Warsaw (I think I heard you are from Warsaw).
Super video! Exactly what I needed! Thanks
No probs!
Wow! Do you have a specific video to show step by stem of how to unmix multiple voices using spectralayer
Same you did not tried out the unmix choir from vocal tool...
I go all the way back to version 7 pro. I use it along with izotope RX 11 advanced and lots of plugins for forensics. For heavy duty buried voices- Hand carving using a spectral scope and manual brush is still the only answer, specially when the voice is only a distant part of the signal.
Only issues so far is the GPU (4090) optimization is useless. Supposed to be fixed in a patch…
very good video
regarding ( add this as plugins in Nuendo )
I think if they take a new path will be even better
like enabling direct offline processing, and add these modules inside proceses
will be something huge
as most of these modules are not made to run in real time
and they can also improve the experience of direct offline processing to make it easier to work with
agree! separate Plugins would be much more easier and faster to use ;-) but always Specral Layers is getting things done in an amazing way!
I've actually liked better the SL dereverb than Acon's. The voice sounded fuller and I think a second pass with Acon or Clear (also great for denoise/dereverb) it actually may get almost into studio quality territory.
Great video, very entertaining.
I must agree with you. Acon still holds the stand as the best dereverb. Both Izotope and Stenibug should buy his algorithm
I did not have good results with dereverb in Spectral Layers…they need to improve this one
Hi there, can you explain your request for destructive editing please? Surely the result would be the same, because that’s the best separation it can do, with the only difference being that whatever it’s decided becomes permanent. With non-destructive, it makes its decision but you have the option to move things manually from one layer to another because everything’s still there. Or am I missing something? Would the destructive method allow for a more powerful algorithm?
That was my thinking too. Destructive vs non-destructive would simply have less storage use, but the effect of competent separation really has to do with underlying algorithmic assignment of the many tens of thousands of spectral "pieces" to their role ("bass", "drums", "vocal", "synth", etc), all of which is done using a set of computer "tricks" or algos (a bit like doing the square root, or doing a sort, only this is likely far more complex and is operated upon the underlying detailed sound analysis, which, oddly, I understand comes out of naval SONAR tech). I did some programming, and we developed algos to handling data the way we had to process it (mine was financial data). Here we know the start is a controlled breakdown using Spectral Analysis (there are menus choosing the exact settings for this) or to call it by its Comp Sci term a window of Fast Fourier Transform analysis (FFT) which then is uniquely identified piece by piece and generates this enormous number of fragments. A big part of assignment would be same frequency, but that is not the be-all, end-all. A singer may be singing a held note of the same frequency as a flute playing along. What probably counts more is start of note and then related hold of it and release, and identifying transients and formants by type. Then assigning that a tentative assignment as a particular instrument or vocal.
In this case the components of the sound are like strings of items numbering in large numbers and getting assigned to categories. Once assigned we have these "lanes" we can listen to.
all that I need, is just to separate from audio, the vocals and piano, into midi, is there anything, that's accurate? maybe chord changes would help.
I'd go for Melodyne!
Would Spectralayers be a good drum removal tool ?
Hi, In this video are you running specctra layer as a standalone or as an ARA extension?
Helllooo .how can i get rit of phasing of drums after unmix . In spectralayers greeting to poland from Croatia.
Live recordings which were recorded under non optimal circumstances are more difficult to unmix
Ran unmix crowd, music, instruments including drum kit elements (Kick, Snare, Cymbals and Toms) on a 40 year old bootleg. Not perfect but very, very good. The next step will be to try the 'transfer' function where you find, say, a lead vocal that is in the 'Crowd' stem and graphically lasso it and transfer it to the Vocals track. No so quick, but should sharpen the stems up enough to basically remix a dirty, tape 'master' recording into something more balanced and listenable.
A great leap for Spectralayers, and the next version should see them out of sight of the competition.
Cool Vid Paul, thank you.
Thanks, Stevo!
This is the next thing I want to learn too. I have stems with Ella Fitzgerald's singing and her back up band (on the "Songbook" albums, say the tune "Yesterdays"), but a held note from Ella was so steady the algo seemed to think this was a flute, and put it into the other layer. So, getting Ella's held note back in the vocal would be nice.
@@FoliaSound As a sometime programmer, I'd think destructive vs non-destructive is not really going to guarantee far better inclusion/exclusion. Instead I think what this entire product design is winds up being a mass of FFT analysis is done on the source, then it creates a giant array or similar store of "pieces" of spectral info, which is categorized as to which layer. It is this categorizing algorithm which we musicians sense. We know what content goes into a bass part, or a vocal. The algo would have to use what we might call "tricks" to assign fragments into categories. An example of an algo or trick would be formant type human voice sounds - our speaking and singing starts and ends often with formant type items from our mouth parts, so the pieces with those are going to get assigned to vocal (or speech, if that choice is made), then the algo would look at the sequential material right next in time and at the same frequency as this obvious "formant" stuff, and assign that to "vocal", and so forth. That's a highly simplified way of describing the algorithm, but it has to be something akin to that. A set of rules used to assign "type" to the broken down spectral content which is now separated into thousands of pieces. This would be the same regardless of if it is all stored separately or if the original file is wiped out and separate track files are made. Today we have so much storage around that it's not a big deal to just hang on to vast amounts of underlying data in most cases. There are exceptions, but it is true there are now often Terabytes of places to store data.
We are now in our post pro studios completely on Spectralayers. Way better results and great workflow for audiorestauration too.
RIP izotope RX.
SL 11 highly recommend!
Nice to hear that!
I agree, at least with denoising. A year ago, almost all my denoising went through RX, but now Clear, Acon, dxRevive and SL took the mantle. The only aspect RX still is king to me is spectral editing and vertical cleaning, and some niche tools like de-wind. Exciting times for voice processing!
great show my friend
Dzięki za super video as always :) czy u Ciebie w trybie extreme tez to trwało wieczność ? :) (u mnie na i9 64Gb RAM iMac ładnych pare minut...)
Ryzen 7950X, dość szybko poszło!
Thank you!
Not sure how you got unmixing of multiple voices to work but I've no success with it at all. It's the reason I upgraded to this version and it does not work on any songs I've tried (after separating out the vocals and chorus first of course). Nice video though.
Here in New Zealand, "packed up" means broken, or not working 😂
Apple Compatible?
Yep, there is a Mac installer, like for all Steinberg products?
@@FoliaSound All Steinberg releases are both Mac and PC
Spectral layer literally destroys izotope RX
R.I.P. RX
Great video! Do you think audio post Production has a bright future? Still? I doubt that:)
Well, you took perfectly recorded voice recording and mixed it with noise and stuff…. Thats not how it goes, and those tests are cool, but not that realistic….
Still, spectralayers is A+
Less words would have been fine...
so nonsens....you take take 2 voices , let them see in a mix and then unmixed.......... what did happend in between , how did you unmix them , what was the proces?
i would definitely NOT recommend buying this. the unmix module can't even separate out the most basic stems. if you try to unmix a basic pop song, spectra won't be able to discern between a kick drum and a snare drum. it's just a giant unusable disaster - same as all unmix algorithms before it. nothing groundbreaking here unfortunately, and now i spent 300 bucks on this useless software because there was no trial... feels like that was probably by design. if you're looking for a way to separate stems from a song, this aint it.
After 3 months, assuming you've had time to test it further, do you still think the same?
It's a BULLSHIT toy for all those todays wanna-be "producer". Real producers don't need something like that. Or do you see Snoop, Dr. Dre, Quincy Jones, Rick Rubin or ANY real producer using that shit? Of course not.
Hey, have you heard anything about a thing called audio post production for multimedia? Have you ever thought, that working in a DAW is not only about producing music, but much more?
Of course this is NOT mainly aimed at normal music producers but post audio production which is a different area.
Quiet down with your ignorance. Flew wayyyy over your head.
~70 Euros for the upgrade is too much.
Considering the number of new features and improvements in this upgrade, it's actually cheap !
RX 11 Advanced upgrade cost from any previous RX Advanced or Post Suite - 329 EUR :)
@@FoliaSound sonnox restore bundle for 100 bucks on sale
@@FoliaSound
Link?