Like the title says. I have the VST3 FX module in a Modules Chain. Then I needed to apply just a single VST3 FX, so I open up the VST3 FX Module from the right panel and try to change the VST, SL11 hangs hard, had force quit…
Win 10, SL11 (all latest)
1 Like
Patch 3 coming next week fix several issues with the VST3 Effects module.
3 Likes
Hi Robin
thank you so much for your prompt answer announcing patchs which is good if improve SL11.
To be honest with you I start having regrets to have bought SL and more Steinberg product.
I am not a professional and for me make music is just a pleasure but I have a feeling to have wasted money.
If was only money wasn’t really an issue but what is dramatic I am loosing my best opinion on this SL and the people they presented with a big show this product.
I really hope that the next patches allow me to un-mix classical music stems and pop music in a good way.
What I ask you , being Steinberg employees, if I will get what I expect from SL.
What I expect is only two words YES or NOT
Thank you so much
P.S. I am really disappointed 
Yes, this is also my impression of SL in its actual state. There are so many bugs, that make working with SL adventurous. I do love SL for its ideas and possibilities, but it seems to be still in a state of development, concerning the stability. I use SL a lot, but also use Audacity for some tasks, that I can not find SL to be reliable enough.
Take for instance a rather simple task like “Normalize”
When doing this in SL
and compare the result of Audacity
The outcome of both normalizations are not the same. I do expect the result exactly the same. This is just rescaling the amplitude by a constant factor, which is numerically easy. And it takes a lot more time for this than Audacity. What is SL doing here?
And so, yes, I do have to switch to Audacity just for a simple task like normalization. I find this unpleasing and unsatisfactory.
This is just one example, which I consider SL not reliable enough.
I also tried normalization it in Wavelab.
→ Three programs, three (slightly) different results 
Which to trust?
So, it might be a general problem… maybe “float” is just not the right format for integer values… but that’s a completely different topic and far far beyond the scope of SL here. In this context, I wonder why 32-bit float has been chosen and not 32-bit integer as a format to record. (Year, it’s for large numbers, ok. Not the fault of SL, WL, AU or any other software). Doing arithmetic in float has always some special flavor in it…
See here to get an easy headache on a nice sunday morning:
https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html
Arithmetic in float feels like the introduction of randomness in the lower bits.
Working with different programs for digital soundmanipulation seems like working with different analog devices. Nothing has changed; the outcome is always different.
1 Like
unfortunate that you guys are unhappy with SL
I’m the opposite, as I’m thrilled with the results I’m getting with SL11…But, I am not unmixing music or comparing SL against itself or other softwares;
I’m cleaning up on location recordings and the results are spectacular in SL11
I have done some music unmixing and manual editing and there is a way to go until SL models will be capable of what you guys seem to be expecting
Ok, to clarify, I am definetely not unhappy with SL. It’s one of those great programs, that changes the whole workflow of certain things, i. e. soundworks.
The idea of layers is ingenious. I love SL. The only things that are cumbersome are the bugs and the imperfect implementation (from some numerical or algorithmic point of view) of some of the algorithms. But sure, what is perfect? No program is and it needs years or even decades to become (almost) fully bullet proof. Spectral Layers, as it is today, is certainly far from this. But, yes, it’s already very well usable in many cases. This program has my deepest respect - a masterpiece of art it is. - and actually quite easy to use, right from the beginning. And at the same time has a depth, that we might not have reached yet. Not every program can say this.
1 Like
Yes that’s true to remove some noises is good
Seems to me that we don’t use SL on the same way.
However, on this case, all modules take only space on my hard disk!!
When we say un-mix a song and give Piano, Guitare, bass drums etc. stems then what I expect is, all instrument are well un-mixed and on my case after un-mixed, stems sounds really bad.
Could be also I don’t have enough experience to use SL correctly!!
Piano mixed with strings, guitare mixed with drums, other, mixed with all instrument etc.
I can post un example which would be better explanation on what I receive from un-mix song.
If we have tips on how to un-mix a song correctly with SL, welcome, I am ready to use them.
1 Like
Well, I guess, unmixing will probably never be perfect in the sense it was before mixing a file up. Just imagine one point in the spectrum, that has for instance 60 dB. Further think about several instruments that have certain overlaps in their own spectrum. And also every real instrument differs slightly from each other of similar kind. So, one can never be 100% sure, how the 60 dB come around. It’s also a matter of entropy, I imagine (just think of the amount of energy necessary for unmixing; AI is computationaly very expensive). Unmixing colors or chemical compounds is far more difficult, than just putting them together.
What SL (and other programs) can do today in this manner is quite good, imho. But it’ll never be really 100% perfect in a mathematical sense. My point of view here is a bit technical with math and physics in the background. Since I do enjoy music very much, SL is a great tool to discover music and sound in general in a new way. Just like Google Earth, maybe… discover the world new.
And yes, there will always be much, much room for improvement. We live in such exciting times with so many opportunities to discover the world around us - isn’t that beautiful?
I am used to spectrograms for about a year now and start “to see sound as spectrograms”.
There could be a quizshow (for nerds), which shows a spectrogram and 5 sounds to chose from. Now, match the spectrogram to the sound. or vice versa.^
The program SL should just run completely as expected, smoothly and without any unexpected behaviour or crashes
1 Like
/* =======================================================================================================
- Script Name: normalize
- Description: easy reduce volume
- ==================================================================================================== */
using System;
using System.IO;
using System.Windows.Forms;
using SoundForge;
public class EntryPoint {
public void Begin(IScriptableApp app) {
app.CurrentFile.DoEffect(“Normalize”, “[Sys] Normalize RMS to -16 dB (music)”, app.ActiveWindow.EditRegion, EffectOptions.EffectOnly);
app.CurrentFile.WaitForDoneOrCancel();
app.DoMenuAndWait(“File.Save”, false);
}
public void FromSoundForge(IScriptableApp app) {
ForgeApp = app; //execution begins here
app.SetStatusText(String.Format(“Script ‘{0}’ is running.”, Script.Name));
Begin(app);
app.SetStatusText(String.Format(“Script ‘{0}’ is done.”, Script.Name));
}
public static IScriptableApp ForgeApp = null;
}
my 2 cents
normalize using Sound Forge? You are saying @Unmixing has a similar routine with SL?
I have had SL 9, 10, 11
9 Was complicate
10 Was more complicate
11 Is more complicate then 10
None in between was possible to have a result I was expecting.
Reading all post each one doing different things, seems to me, navigate in an ocean trying to reach the ground swimming.
@Robin sorry, could you please takes time to prepare a form step by step to obtain an un-mix correct?
Do I have to open a ticket to obtain it? (already opened 2 tickets without good answers)
I can see that now become more complicate then before
Thanks
@Gennaro Reading your posts it seems you may be misunderstanding, or expecting too much from the state of the art AI technologies.
AI (deep learning) is not, and will never be as accurate as a math formula. But no math formula can unmix songs. Actually, AI is closer to cooking experiments that math. And it will always be approximate, to a certain degree, depending on the kind of materials it has been trained on, and the kind of material you submit to it.
Still, it’s the only and best path we have right now to automated unmixing. And it only got better and better with SL9, SL10, and now SL11 - and at the same time (as opposed to what you say) it got easier and easier to do with each version, SL9, SL10 and SL11.
SL9 required you to go to an Unmix menu, then click Unmix Song, then adjust parameters, then click Ok.
SL10 required you to go to an Unmix menu, then click Unmix Song, then click Ok.
SL11 requires you to click Unmix Song, and click Apply.
Each new SL version required one less click than the previous one, while offering higher quality and more instruments than the previous one, so I really don’t understand how it got “more complicated” ?
Try unmixing the song samples from here and see if you get the same results from SL11 as can be heard on that page.
If not, then please create a new topic about this to report it.
If yes, then everything is working as expected on your computer, and it may just be that you are submitting songs that the AI has more troubles unmixing, because it was less trained on this kind of recordings, or because it’s harder to identify specific timbre. But keep in mind it’s still state of the art in 2024, it’s highly unlikely you will find a better automated technology to unmix them right now.
And then, there are the deeper layers of SpectraLayers, the spectral tools which allow you to fix any mistake from the AI, and get the perfect unmixing. That’s actually the core of SpectraLayers technology, but it takes a certain learning curve to master. A preview of how to use these tools to fix unmixing results can be seen in that video :
Note that even that aspect got easier with SL11, since SL11 introduces the new Transfer tool.
Hope that helps.
4 Likes
Thank you so much Robin
become now more clear that I was confused
I will look at all example and maybe I will be out from this un-mix process better then the actual.
Good morning Robin
thank you so much for this video that I have seen several times.
It’s true that I was requesting more then what SL is able to do.
Most likely classical music and particularly strings in this video are fantastic and so, I don’t understand why my strings un-mix not in the same way.
Saying that what is missing in my knowledge as well VST which I didn’t find the same, how I have to adjust then the stem after an un-mix.
In Steinberg VST I have only few and not all, could be a problem to setup preferences.
I heard also all stems and the part VS with other systems you posted.
Surely SL is the best at the moment I agree with you.
Am sure also that the un-mix process will be better and better with time.
.
1 Like