I know that the tools now have had lots of attention
in v13, which I thing have been a long term wish for many people, including myself.
I guess the tools a getting closer or better that example the iZotope RX tool now. But does that mean, that the unmix song, the separation stuff has been left alone in v13?
I think that v12 wasn’t really good with bass stuff. In fact I think separations / instruments was all over the place on the separated tracks.
Bass stuff was hard to do, since the deep stuff in a plug bass wha placed on the bass track, and the other more clicking / plugging sound in the bass, was placed on different tracks. I think the bass separation in general, is worse than on earlier Spectralayers versions.
For example was the guitar separation often used for other instruments, and most of the synth stuff in songs, was placed on the “other” tracks.
How is it in v13? Maybe the new tools are making it easier splitting up more stuff on the tracks. “separate instruments” seems to have been improved. I never thought it worked as it should in v12.
Nothing changed in regards to unmixing songs in version 13.
If you want state-of-the-art unmixing with a lot of different stem types you either can use Mvsep.com (very easy to use and free, if you’re okay with waiting for processing) or, if you prefer local usage, you need to set up a Python environment and ZFTurbo Music-Source-Separation-Training and load, for example, the MVSep Mega 53 Stems model.
Looking at mvsep.com, the “free” version is somewhat limited: 16-bit WAVs, 100 MB max. PRICING PLANS:
Their paid plan is credit-based, and looks pretty cheap (a few cents per minute of program.) If it works well, that’s probably not a bad option for most people. In my case, I tend to unmix entire sets. I could break sets into smaller song pieces, but I’d rather run the unmixing locally.
It caims to use 100 different AI processes or algorithms in combination. Seems pretty elaborate. ALGORITHMS:
Except the exclusive Mvsep models, all other models offered on mvsep.com are also available for download on Huggingface and will run locally, either in Ultimate Vocal Remover or in ZFTurbo MSST. Many models can also be directly downloaded inside the program settings of Ultimate Vocal Remover, so you don’t need to search for them.