Transient Tonal Splitter Model Idea

[Nexeration]

If you familiar with music production maybe you came accross the plugin Split EQ which seperates Transient and Tonal Part.
The weakness which comes with Split EQ is that if you extract tonal and transient split and put it together in a null test it's not completely clean and there are a looot of artifacts.

So one method would be to create either an algorithm which focus on transient or one which focus on tonal.

It would highly effective to have such a model how does that for vocals or individual sounds like drums. The model UVR-De-Echo or De-Reverb is good to begin with but not exactly the transient-tonal splitting I mentioned. Would be nice for a lot of producers.