With that, I finally make a statement here: SL 12 in its actual version is just crap, a bunch full of bugs! This programm, as it stands today, is unusable and completely unreliable!
And I paid for it…
Nice buzzwords: “Preserve Formant”, “Loudness Countour” → what do those options exactly do? Anyone here uses it?
Because I know it! When changing the time, one either raises or lowers the frequencies by changing the speed (when changing the playing speed as on a tape recorder for instance. No sample points are added or deleted. Not talking about fourierstuff here)
Is this a serious question?
I am not a Spectral Layers fanboy anymore!
→ " Not talking about fourierstuff here" We just stay in the time domain, when using the proportional option. Or do we not? Only enter the frequency domain, when necessary. Proportional timestretching does not require frequency domain. If SL goes nevertheless into frequency domain, because the dataset is always spectral, than … fine. But I would consider this as unnecessary and in principle a slight mistake, because the transformation back and forth introduces numerical errors, which are avoidable here.
Hold on… is this any ‘broken behaviour’ here necessarily.? I mean, can you check against previous editions (SL10, SL11, etc..?). i.e. has this ever worked as you expect..?
Do you have access to WaveLab Pro.? Can you check against behaviour in its spectrum editor there (same material, same procedure).?
Check with Audacity for instance. Yes!
Sure, Wavelab can be used, which I also have.
Sometimes, I doublecheck Wavelab with Audacity also…
The Steinberg universe does not seem as reliable as I would expect (but that might be my case only, though).
I do expect exact results in a numerical sense and also all functions/moduls work as they suggest. This is not the case - here and very often elsewhere.
This can and should be critized openly and direct, which I do often. Not to the likings of some, but that’s the way it is.
There are some things, that just have to work without any question, imho! Those are the rather technical simple tasks. This breaks so often here. As can be seen by lots of other comments.
Steinberg should work on quality first, before implementing more and more features. But this might also be owned by the hard marketing competition. So, year, we are talking about software made for artists, not scientists- which is fine. But it should be stated clearly.
Of course, criticising is fine when done politely and respectfully - along with clear evidence to support the claim (as you have provided).
Still, my question remains - has it always been like this, can you check against SL10, SL11, etc..? (uninstall SL12, download and re-install SL11 for example.?)
When slowing down proportionaly, it seems to work ok. (But this I now must double check also!)
No, I do not usually check with earlier versions, because I expect the software to work, because I assume, that the manufacturer makes thouroughly tests before releasing, because they have a certain reputation…
This - has failed!
But here is an appraisal: This forum is open to the internet and without almost any censorship. This deserves an honest appraisal for sure.
It would be nice, to hear something from Steinberg about the actual version of SL and when they are planning a fixed release. Haven’t heard about anything in this direction, yet.
I remember this happening to me too in previous SL versions @Sunnyman . I typically use the transform tool instead to make sped-up remixes based on pitch as demonstrated here:
The Preserve Format option isn’t factored in when performing proportional shifting. Enabling loudness contour tries to compensate loudness differences after shifting audio.
I am a customer, not a paid beta-tester !
It’s a pain, that every serious result must be doublechecked with another program.
I do not have any trust left in Spectral Layers! As it is - right now (v12.0.0) - this program is not recommendable.
Same thing Elastique does over the past 20 years when preserve formant is clicked in for, say a female voice….to then dial in the center freq for that particular voice to keep the sung words from getting chipmunk-like at extreme speed changes (with or without accompanying pitch up/down chosen in conjunction with speed).
I’m sure you already know that though. It’s not for every signal on the face of the planet.
Personally, I don’t use spectralayers for speed/pitch tasks.
Well, that’s not the point, imho: “Personally, I don’t use spectralayers for speed/pitch tasks.” Fine, though, if you don’t use it.
First: I use it very often to transform frequencies above 20 kHz to audible.
Proportional timestretch is, which is just a slower playspeed in general, fits my needs perfectly.
Second: SL has this option. Is it too much to expect a technically rather simple function to work 100% clean? If it can not be done 100%, than pls remove it, until it works 100%.
If such simple stuff like proportional timestretch (and the signal generator as another example for simple stuff) does not work properly, how can one - honestly - expect all other, far more sophisticated stuff, work properly?
I do consider this as in someway rediculuos, that this has to be mentioned at all - for a software of this caliber. Some say, it’s the cutting edge of what’s possible today - and i tend to agree. SL is ingenious conceptually, but very poorly from a technical point of view, imho. (Another example for this might be the slow down, when using the magic wand selection tool for lots of data. It behaves like O(n^2) runtime. For a mere selection of datapoints (!) - seriously??? In the year 2025???)
