After several month using SL, I still like this program, but…
and this is a very big “but” !
This program is so buggy, that I use SL only with high caution on important projects.
There are so many bugs and inconsistancies, that I have stopped counting them.
The program seems not be very efficient from many algorithmic points, it is often very slow, even on very fast machines.
Compare with Audacity or Wavelab for certain parts. Spectral Layers is tediously slow.
Maybe there are many lines of old code, that should be rewritten completely.
The design and the ideas of SL are great and without any doubt ingenious. The actual implementations seems like a developer version, that is still unfinished in many ways.
My wish for SL 12 is therefore a debugging with high priority and please much more efficient algorithms overall. We need more parallelization where possible and more use of GPU.
Why, for instance, does it take such a long time to copy a selected intervall into a new project?
Why does normalization take so long?
Why does the Magic Wand selection breaks up so often or even crashes the program?
The algorithms used do not seem highly efficient. Maybe the mappings of the spectral data need a redesign?
The use of memory could be more excessive, imho. 32, 64 or 128 GB RAM are nothing unusual or exotic today.
I know, this is not something many want to hear, but I do not want workarounds for this or that. If it is in the menue, it has to work, without any workarounds and at best, just as expected.
I’m not experiencing the buggy behaviour you outline here
SL11 is capable of amazing things
IMO, many of the users here are looking for automated processes…The best part of SL is the manual tools…alongside a few exceptional modules…like Unmix Noisy Speech
When checking “Transfer”, the mouse moves very slow and jumps in ca. 0.5 seconds steps. This tool is unusable.
Why is “Normalize” so slow? (I use it very often on files larger than 1 GB and there, it makes a difference… Wavelab/Audacity are around 10x faster(!), Multiplication of a dataset by a scalar is fairly easy and could be done parallel.)
This issue with the Transfer tool was indeed reported by a couple users but I was never able to repro it locally with any machine so far… That being said you can simply disable realtime preview of the tools by going to Edit > Preferences > System > Realtime Brush Preview Limit > Set to 0.
Normalize is slow because it operates on spectral data instead of waveform data (like all tools and modules in SpectraLayers), but in that case I suppose spectral data is indeed unnecessary as you usually normalize time ranges and not specific frequency ranges, it could be optimized to be much faster.
-Is it slow as soon as you move the cursor hover the spectrogram, or when you start clicking inside the spectrogram ?
-Is it slow if you zoom very closely in the spectrogram first ?
It is slow, once “Transfer” is clicked and moving above the spectrogramm. It’s “jumpy” rather than slow; hence unusable, feels like stutter step through oel.
Btw, what does the cache have to do with it? Why not use RAM in any case?
32, 64, 128 GB or even more RAM is nothing special today.
And yes, I do not want lag.
The VST-plugin-interface still seems buggy or not very efficient.
um, clearly I participate quite a bit here on the SL forum
and yes there are buggy situations, like the latest update
but, I have been a part of situations in the broadcast profession that were far more buggy than SL! Software in which many hours of human effort down the tubes for no good reason
Proteus and Clarity to name a couple
SL is capable of amazing things…I’m just not one who leans too heavily on the music side nor many automated modules beyond Unmix Noisy Speech
I tried again with multiple Windows computers, even old ones (cheap laptops from 7 years ago) with multiple projects and I was unable to repro the transfer tool issue you have.
Would you be able to do a video recording of it, with the full screen visible ?
