Short selections not responding as expected, small linear fade in/out?

Just trying to do something very simple
Read through the bits in the manual but Im a bit stumped…
You see here I have created a layer (conga downbeat) and moved it to the beginning of the project
I can see any zero crossing options so the selection is just cutting to the exacting selection window
Now I need to remove the click with a very quick fade
I have selected the required rise time length
If I cut, with no fade applied, nothing happens
If I erase with full hardness nothings happens

Is there a tutorial video somewhere for all the basics especially when just wanting to use as basic editor (as well as spectral) in the one tool?

Zoomed In

Zoomed Out

It seems this was an earlier bug…could anyone confirm that eg a small 5ms selection will cut/delete the contents?
Also how do you do a simple linear fade in in the above application then?

Firstly - I do not claim to have extensive knowledge about how spectrals, FFT and connected stuff is calculated or how it is actually programmed in SpectraLayers, I just speak out of my own experiences, observations and tests with SL.

As for how it works right now, personally I would not recommend SpectraLayers as a general basic editor (cut, crop, copy, paste, fades … whatever), at least not when you need absolute precision. If you have to, though, you should pay close attention.

SL mostly works in spectral domain with many processes dependant on the FFT Size you can set in the Display panel. With that setting you can shift the focus to either higher time precision (lower FFT) or higher frequency precision (higher FFT), and with the Resolution you can additionally refine the display (x1-x3 are just for time resolution, x4 adds one step of frequency resolution).

I assume SL calculates an image for the display with a certain size in pixels, depending on the sample rate of your project and the set FFT Size and Resolution. And there start some problems you probably encountered - the pixel resolution varies according to these values and it only matches a sample or time position at certain intervals. As most of the tools (like the Eraser, but also Cut or Delete operation from a selection) edit the underlying image in pixels and not the audio samples, there will be some kind of quantization or shift.

Example 1:


A short 48kHz mono file with a 0dBFS constant signal, just for better understanding.
Display is set to FFT Size 128 with Resolution x1.
I do a time range selection (no time fade) of 200 samples and hit [Delete] on the keyboard or use EditDelete.


The result is not what I expected. There is some fading and also a general shift compared to the original selection.


When I change the time ruler unit to Pixels I can see that the result matches perfectly with the grid here, though.


I revert the edit, keep the same selection but change the Resolution in the Display panel from x1 to x3 and repeat the delete process. The actual pixel count changes by factor 4, which allows for a more accurate application of the delete process, but it is still not exactly where I intended it to be. The result matches again with pixel grid.

At 48kHz using FFT 128 at x1 results in 32 samples per pixel (or ~0.667ms) , at x3 it refines to 8 samples per pixel (or ~0.167ms).
In your example screenshot you also use 48kHz but with FFT 3072 (and probably at x2 or x3, whatever the default is). At x1 you have 768 samples per pixel (or 16ms), at x3 it is 192 samples per pixel (or 4ms). Considering you have selected about 2ms (maybe even just below half of 4ms) I assume your selection end point was quantized/rounded to position 0, so no wonder there was nothing happening.


So much for spectral based edits.

Under Edit you also have the options Insert Time, Delete Time and Crop Time which allow for precise sharp edits like in a regular editor, but they also suffer from quantization/rounding.

Example 2:


Again a short 48kHz mono file with a 480Hz sine wave. I select exactly 3 periods of the wave (300 samples) and do EditDelete Time.


Result … yeah … no. There should be no spike in there.


Changing the time ruler unit from Decimal to Samples and doing Delete Time again I get a clean result with a continuous sine wave.

To be fair, I can use Decimal or another kind of unit and get the correct result, but I have to manually change the decimals for the start and length. This is nothing any person would likely do, so here is definitely room for improvement. I would be fine with a rounded displayed value, but for this kind of function it should use the actual selected range in samples for the calculation (just like it works in Excel and variants, for example). Please have a look at this @Robin_Lobel.


Regarding fades:
There is currently no one-click solution for easy fades and crossfades.
The more cumbersome old way is to delete some of the audio with a certain desired time fade.
The more elegant way is to use a layer’s Envelope. This is new to SpectraLayers 11 and for now pretty basic without automatic options or curve smoothing, but it works and you can adjust it when needed.

Maybe that short overview helps you. :sweat_smile:

3 Likes

@Laturec
Thanks SO much for this overview

BUT
Its an operational issue. Most of the time Im using as an ARA in Cubase. This means as soon as I start SL extension, I no longer have access to the sample editor, which is very basic at best and still have to open wavelab for anything of substance so then I have to workaround, bounce a copy etc and its all just messy.

This is especially true as I start to use SL as a very creative tool for stuff I wouldnt even dream of previously

My hope would have been simply to be able to use the waveform display and standard Steinberg shortcuts and paradigms to condense the workflow.

I dont do a lot but just depends on client and creative scope allowed…but just the basic sample accurate range in waveform view with some cut/copy/insert/overwrite would do the job and the new volume automation with curves

Cheers

1 Like

Thx for the overview.
Some of those issues are not new, though…

It’s a very good question to ask of how the pixels on the screen at some resolution correspond to the actual acoustic data.

Just imagine one selects some points and than changes the FFT-size or the resolution. How does this effects the actual acoustical datapoints?

See here for instance:




Not forget about the fancy 3D-view… any real good usecase for this?

I do love the concept of SL and consider as one of those great ingenious programs, that combine art and usability. But the actual code base and implementation of SL seems to need some deep refinement on a very fundamental level.
As it stands today, I consider SL as a nice artistic toy, but not good enough for precise work in a scientific sense (which might not be the intended primary use, though).

I might be a bit sarcastic… sorry

I have needed to change the 3D angle on occasion, but maybe 3 times I’ve used it over 10s of thousands of edits…but when you need it, it is very helpful

that said, I do not recall my specific use case

When you have an ARA extension open in Cubase you can’t open the sample editor because the ARA extension will be the actual editor. This goes for all ARA extensions, not only SpectraLayers. You’ll have to leave the ARA extension through “Make extension permanent” or “Remove extension from selected event” to be able to open the Sample editor in Cubase.
This is the way the host (Cubase, Nuendo) has implemented the ARA protocol and I don’t think it is an issue with the ARA extension/SpectraLayers per se.

Try to ask at the Cubase forum, maybe someone there could help, e.g. if there are any settings to change in Cubase (I’m not aware of any though).

Thanks @Marshall
I understood those limitations and highlighted, that for the simple audio editing that is lost when SL becomes the"owner", it would have been nice to be able to do the basic audio stuff in the waveform editor.

Without it, there is no easy workflow and everything needs to be baked in order to allow a diff editor back into the workflow.

It is what it is…

Cheers

1 Like

My guess is that @Laturec angle and examples are spot on, on the situation at work in this paradox. This is not a dilemma between an artistic or a mathematical tool. Some of the many Spectralayers’ features are the most exact one may find in the demixing software field, including the recombination of a file by summing the stems resulting from its demixing process.

However, the following may baffle some readers but spectral editors utilization of Fast Fourier Transformation (FFT) technology by itself (not something developers may avoid) dictates a choice of frequency domain precision OR time domain exactness. It is simply not possible to attain both at the same time.

This subject has been treated in this forum in several ways, probably the most complete being this thread where @Raymondo shared this excellent article explaining FFT and windowing. I hope those interested or just curious may have a look at it and divulge in case deem it so, because this is still a novel field for many.

About the dire practical implications of this limitation, my response has been very early on to keep using both; an extreme sample-level precision time editor; Sound Forge Pro (I understand Wavelav and others might be close to its level), where I process all those time exact editions. And that tool together with Spectralayers (plus Melodyne and other) Frequency domain editors & effects …While finally sum it all at a precise and able 64bit floating point engine DAW, Samplitude X.

3 Likes

I am glad I read this… It is an important conclusion.

SL is a SPECTRAL editing tool with AI-driven audio processing capabilities. It is NOT an audio editor in the accepted sense (WaveLab, Audacity, etc…).

SL will not be your ‘one-stop’ solution for all eventualities - however much that may be desired. As @Nspace rightly points out, we still (I still.!) need to rely on several ‘tools’ to get jobs done. Nothing wrong with that.!

However, yes - I agree from reading several other posts this lazy Sunday morning over breakfast, that folk such as @Sunnyman and @Laturec have pointed out very real ‘limitations’ and indeed ‘bugs’ that I’m sure @Robin_Lobel will respond to, given time. And its clear there are also (thank you @Laturec) many areas of improvement for future attention around some of those ‘basic audio/editing needs’ that may eventually make life a bit easier, maybe reduce the amount of back-and-forth between apps/tools.

The point of my post here, was responding to something the OP of this thread illustrated - namely, a need to reset user expectations (with specific regard to ‘basic editing tasks’ and what SL should be providing). Although these type of functions exist in SL, they are NOT at a level of a dedicated app (which was never SL’s intended use-case), and perhaps never will be.!

In summary, a multi-pronged approach to sound-design, editing, processing, mixing, mastering, management, workflow, delivery, etc… will (continue to) be needed.

[EDIT - off topic slightly; I found this an excellent post, thanks.
Short selections not responding as expected, small linear fade in/out? - SpectraLayers - Steinberg Forums]

2 Likes

@Puma0382

Great response on many levels

It does require a reset but that depends on context as well

I use (was hoping for) sl would be for sound design and bleed editing etc as i do a lot of live group recording inc drums

I originally did a lot of the ( basic tool/sample accurate editing because timing is important in music) in wavelab which doesnt have ara but is reasonably integrated. My hope was the spectral unmix functions would add great help with acoustic singer clients. Its taken me a while and the last job was satisfying and prob best i have done using r88/f8 nulls and then unmixing to control balances, add ms to single elements etc although it was pretty decent to start with. It allowed some things that couldnt be done in reality but end result was a very connected and sweet performance. Sl only needs a couple of things to actually be a successful composite tool (for my application) as it already has a waveform window

1.sample accurate selection with fades real
2. Phase rotation

I have been checking out other options now and rx pretty much does what i need ie just those couple of extra tools and both rx and acustica have solid phase rotation tools. I actually found acustica is prob the go to so far but is a bit buggy.

Unfortunately, the mentioned items are quite essential to my workflow and ill make a decision soon. Um trying to simplify and have thrown out 1000s of $ of plugins…and options paralysis is just about done…and work is sooo much more pleasurable and spontaneous/fast and this is the nature of right brain success ie muscle memory

I find it interesting that when i asked about phase rotation on cubase forum, only got 1 or 2 responses and a lot that made comments that indicate they have no idea of the pact of assymetry on dynamics response…so i really dont bother much on the forums anymore :frowning:

Couple quality users but nothing like its been eg 20 years ago

2 Likes

Does the Phase Rotation in SL12 not work?

1 Like

Thanks. And yes, absolutely - I did get caught up in ‘generalising’ the situation a bit too much perhaps…

But isn’t it great, there’s so much passion and love for this amazing app among users. And maybe only natural that folk are excited and marvelling at seeing what it provides and then find they ‘need it’ to do just that little bit more… And it would be ‘perfect’.!

Great to read the outline of your specific use-case and the solutions you’ve found. We can only wait and see what plans @Robin_Lobel has for future editions.

I was going to go on a bit more, but I think enough to leave it there…

PS:- as @ctreitzell has posted, phase rotation is there now in SL12 (I have it, but not installed it yet - will get to it around next maintenance release.!)

I only have 11 and likely not upgrade atm…although i requested phase rotation a couple of times so if its there now thats impressive response :slight_smile:

1 Like

@Aurasphere yep, you asked and SL12 got it included :stuck_out_tongue_winking_eye:

1 Like

Well thats great!
Tbh my income is manufacturing high end acoustics (mainly qrd) and translation speakers as a bespoke thing…and had a work injury so im pretty stuck for income (as well as letting music clients go so i could start up).

Every purchase these days hurts so im also trying to invest in what really matters, not just wants, hence the declutter

Would you believe i still use wl6 and have stuff coming up that is going back to cd redbook…still works…quality sw

Unfortunately sl12 would have to wait a while :frowning:
Acoustica had a special for 49

However phase rotation is a critical tool esp when using vintage mics and tube pres. For vox..they seem to really emphasise

Still…a great response to request although would be great to have a review on phase rot feature, its impact and qualiry etc

I feel you, brother…my budget is just too tight…right now

and SL11 does the majority of my needs…thought I was done with my film, but improved skills have me fixing things I couldn’t conceive to repair from when I started with SL10 in May 2024

I’m hoping for your windfall :slight_smile:

Thanks for the thoughts

Tbh prob phase rotation is a bread and butter tool thats hard to go without.

Def would be good to hear feedback about it :slight_smile:

1 Like

agreed, phase rotation, new envelope design and new harmonic selection along with many other additions to SL at version 12 make it a no-brainer for me…just, waiting for the dust to settle on my finances…Tires gotta get purchased first

The phase rotation module does what it should, as far as I can see/hear.

It does take veeery long time for my Mac Studio M1 64 GB to calculate the result, though. Seemingly as long time as an Unmix song task with a couple of stems ticked would take.

1 Like

Thanks, that is very helpful

1 Like