Who would be the person or e-mail address here if I want to report bugs or to give some ideas? Or just report my experience?
I know it might seem to fast, but I do use various sound to demix and have some experience with reporting bugs in gaming industry. (tester)
If there is an issue which in my opinion should be worked out I would like to help them or even test features. I do understand they might don’t want to talk about it in public but they still would take it seriously with the well chosen line with communications and language. After all it is their interest to have their software work as good as it can be.
SpectraLayers uses different open models from GitHub and Huggingface. It shouldn’t be a problem for them to allow us to load custom models. In tools like Ultimate Vocal Remover or ZFTurbo MSST you simply load a custom model and a config file and it works.
Well @Robin_Lobel has already been tagged earlier in this (somewhat) mega-thread
As far as I understand things, Steinberg is essentially the shop window for Robin’s SpectraLayers ( I believe Robin has said that in the past on this forum)
and can kind of be seen here: https://en.wikipedia.org/wiki/SpectraLayers
Robin customizes the models he imports.
The de-mixing component of SL is already unstable. Until Robin has a stable method for dealing with user’s hardware, it seems like folly to introduce more complexity to the app by allowing users to import models. Won’t happen. After 12 was released, he spent many months almost totally focused on stability issues related purely to unmixing. Some might argue he left that work unfinished. Be careful what you wish for.
Not all models. There was a fallout with a creator of such a model (the drum splitting model) because it was simply integrated 1:1 into SpectraLayers, without any change.
And it’s easy to finetune. You simply can do it in ZFTurbo MSST with your own dataset.
Spectral Layers is a plethora of bugs, glitches and crashes all over the place. An ingenious concept of a program, that needs some very serious recoding/bugfixing and redesign of ineffient internal data structures, imho.
SL is by far the best spectral editor → stem separation is not the main playground for SL, imho.
Nah. A percentage of users would then be yakking about their customLoad models…on Robin’s time…no doubt a sub-percentage of those guys whining about this & that.
