SL Voice Denoise/dereverb vs Final Cut Pro Voice Isolation


Adding SL 9 Pro to my Nuendo tool set. With V9 I think its ‘getting there!’.

However for many of the voice denoise/dereverb tasks I get, nothing beats the Final Cut Pro (X) Voice Isolation process which is truly miraculous. Ambient reverb just disappears!

It’s a real time, one slider deal and produces artifact free results even at extreme settings.

No idea how it’s done, but I hope SL is aware of it and can step up to give this kind of quality in SL.

I literally take voice recordings and run them through FCP just to print that effect to them. For the result, it ends up being quicker and better than any SL/RX/Sound Soap/Accusonus equivalent.

If Apple sold it as a plug in on its own, I’d pay the price of the whole FCP app!

Can’t reveal much at this point but you won’t be disappointed by SL10…


Thanks… reassuring, and something to look forward to!

I’ve gotta say I’m not a fan of how you guys do business regarding updates and upgrades. We see just about zero updates during the life of a version, and here for example, we’re told to wait until the next version to see a better denoise/dereverb algorithm, or in another thread, that the unmixing will be better in the next release. Yeah ? Well why so . It’s not unusual to come across little idiosyncrasies or bugs when working with SL, and I’m happy (kind of) to work around them, but feel my loyalty is never rewarded with any improvements until the next paying version. Sure nobody’s forced to keep buying the updates, and I don’t expect you guys to have the same attitude to updates that Logic Pro has, but a more balanced approach to the "generous/give absolutely nothing " policy spectrum surely wouldn’t send Steinberg broke , would demonstrate some good will towards your faithful and supportive users, and is not something completely unheard of among software companies. IMHO, AFAIK, etc…

There are usually 2 updates post-launch during a SpectraLayers cycle - and that was the case with SL9 as well, although these were just bug-fixing updates this time.

However SL8 introduced native Apple Silicon support for free 5 months after launch (which required a lot of code rewriting), and SL7 introduced many improvements after launch as well.

So it’s not unusual that SL sees improvements post-launch. This time the work on SL10 required so much R&D that it didn’t left a lot of time for SL9 feature improvements.

1 Like

So apart from the Apple Silicon support, there hasn’t been many post-launch updates with noticeable feature improvements since SL 7.0.20 on the 30/09/2020.
Since we’re on the topic of maintenance releases , I’ve noticed 9.0.20 has dropped, with mainly ARA improvements that don’t concern me since there’s no ARA for Logic Pro.
Also, there was again no in-app notification for the update as far as I can tell. ( Yes I can see the irony in whinging about not being notified about an update that I previously said I didn’t have much interest in). :grinning:
Maybe I need to permanently have open one of those Steinberg apps that I’ve amassed, but I’m not sure which one: eLicenser Control Center, Steinberg Activation Manager,Steinberg Download Assistant or Steinberg Library Manager. :crazy_face:

Anyway, I’m looking forward to SL Pro 10, and after checking, have to admit that the major updates are very reasonably priced, and my complaints regarding the content of the post-release updates are maybe not warranted… Cheers Al.

I also look forward to version 10 as well.

From the main developers response, I get the impression that he’s the only one working on Spectralayers (which then defeats the purpose of it being acquired/purchased from Steinberg in the first place). I initially thought that when Spectralayers was acquired by Steinberg that a ton of developers would’ve been assigned to various parts of the code to improve it and work on feature improvements, but now it seems like it is likely one developer only is working on Spectralayers. One thing Steinberg is good at is optimization of their products (like VST’s) and improving their products and that is something that I had initially imagined Steinberg doing with Spectralayers. I’ve talked a lot about the GUI being laggy and using a lot of layers being laggy (with refinement resolution fft size turned up) but I haven’t talked about selections being laggy as well. When you have hundreds (if not thousands) of selections, the app becomes extremely laggy (sometimes non responsive, even on the post powerful AMD chips). So I assumed Steinberg was aware of all of these issues and were diligently working on improveming these issues, however it appears Spectralayers is still mostly still handled by one person (and not a team of software engineers like I had imagined).

FYI feedback about engine optimization has been heard and addressed in SL10.
Would you have a project that shows optimization issues with selections as well ?

Yes, but you can experiment for yourself. Try experimenting by using the magic wand tool and try to continuously select hundreds of selections (Tolerance set at 3.0 db and Maximum Width set at 4.00 S with fft size set to 8192, resolution maxed out and refinement scaled all the way to 100 %).

The only way I am able to get around this is by allocating the selections and saving the selections in parts (for example a series of selections would be saved as “Selections Part 1”), however it becomes an issue when you try to playback what was selected (and because theres hundreds if not thousands of selections) the audio stutters and cpu buffer underuns occur and GUI becomes laggy.

Let me know if you want me to send you a project file demonstrating this.

Something like this ? For now I’m not experiencing any slowdown (macbook air m1). Let me know if you have a project.

Check your messages

Thanks, I can repro a slowdown with your project. I’ll see what can be done, however this really is A LOT of selections.
Selections are usually more a way to transfer data to other layers, or quickly restrict processes to specific areas. After selecting a chunk of voice, why aren’t you transferring it to a separate layer and then move on to a new selection to select the next chunk, transfer it, etc ?

Right! Indeed this is a lot of selections but like I said I usually work around this by allocating those selections (deselecting and breaking them down to smaller chunks) to a “save selections”.

Normally I will work around those selections issues by allocating (breaking them up) those selections into different parts(Saved Selections). However (unlike this very example) there are instances where I need to hear those many selections in order to hear those frequencies properly. For example, lets say I want to unmix an Orchestral performance that has 3 different opera singers (along with a plethora of other orchestral instruments), it’s kind of challenging (non-intuitive) to keep cutting to layer below and reverting back to playback of original audio without entirely hearing what you are cutting exactly (also keep in mind that as you are cutting to layer below and playing that Orchestral performance, it gets difficult deciphering what tonal elements belongs where). If I am able to hear my selections(without audio playback stuttering), I can determine **(more effectively) ** what elements belongs where. So there are instances where it is necessary to make many selections and hear those selections in order to decide whether to cut to layer below (the example I gave wasn’t the best example because the vocal is spread out throughout the song, however there are instances where many instruments/vocals are cramped into a tiny section and you need to be able to hear those selections properly upon playback in order to know if you are cutting the right frequencies)

I hope I didn’t burdern you with more work because really I was just trying to make the point that I assumed a plethora of software developers were working on Spectralayers as opposed to one. If it’s too much work then I would leave it for an update (during the 10th edition)

Something like optimizing selections can easily be achieved because I use Unreal Engine and it seems to be well optimized (even on not the best hardware)

“A plethora of software developers” or just the old idea of the close knit team of developers might not be what you require nowadays, both to reach progress with certain software and also to run a successful firm. And Spectralayers might be a very good example of this. The model corresponds more to an associative one, if I am correct.

Yes, Spectralayers for sure had input at the edition engine, selection tools menus and else editor-related creation from the Sony Creative (Soundforge and VEGAS teams) partnering.
Not for nothing one could see Gary Rebholz (Sony Creative main developer) demonstrating all the Spectralayers minute details and advanced usage for hours at the public released tutorial videos back then.
The same way one could guess that today Steinberg teams assist Robin at VST3 and ARA implementation at least, let alone the obvious input in the registering, license and safety processes, plus marketing and distribution.

That said, I’m glad we’ve witnessed Robin himself crossing from deep imagen-spectral scientific research into technology applications with audio and even early AI research and design. These IMO well above the requirements of the previously mentioned tasks that developing Spectralayers demand (like GUI, edition development, debugging, formats/Codecs, etc) while much more relevant for keeping Spectralayers at the forefront of its field.

So yes, I am glad the software I use and for years trust is based in associative partnerships, presently in very good hands, rather than a traditional in-house team.
Not that I prefer one model per-se over the other, but in this unique tool this alternative has already demonstrated its benefits at large.

True, but you have to keep in mind that there are other competitors in the same spectrum (like RX). That other competitor is able to efficiently do things like optimization and debugging because they allocate their tasks to a team. When one person is doing everything by themselves (although doable) it puts a heavy burden on the person.

I would much rather prefer a team because this would make things so much more efficient. Maybe not for everything but for some parts of coding like debugging or engine optimization. That way the main developer can focus on other things like feature improvements and adding new features.