vocal-remover VS friture

I use this code: https://github.com/tsurumeso/vocal-remover

There is Vocal-Remover with various turnkey GUIs available like this one: https://github.com/Anjok07/ultimatevocalremovergui

I coded my own one, but the general logic and methods that I used were from a paper and a couple github repositories, this being the most useful one I think

I've wanted to mess with Bubble Pop for ages, but unfortunately, like most kpop there are no official or leaked instrumentals/acapellas/stems. Today I started experimenting with a neural network for extracting vocals and instrumentals from mastered audio (vocal-remover). Bubble Pop was the first song I tried it on, and it worked stunningly well. I got a pretty well-separated instrumental and vocal track, and I used those to make this custom mix.

No worries. In case you don't find it, here's a link. It's a GUI version of this one. Both are open source, MIT licence (you can use it commercially). But apart from a GUI it also provides additional models, which are better than the original one from vocal-remover. So even if you don't use the GUI (or can't for technical reasons) you can just use the models from UVR-GUI directly with vocal-remover.

As far as I know, the only more accurate tools for resonance are those that show full spectrograms, for example: In-formant, Friture, and Praat.

In this case it's Friture: https://github.com/tlecomte/friture

Get a spectrogram like this one: https://friture.org/ (use just one graph the 2d spectrogram, dm me for settings) try 2 things: first of all breathe into your mic and try to move the lines/groupings of data upwards and downwards by moving your tongue and throat. Then do the same but take a video of white noise (from youtube) on your phone and place the phone speaker against your lips. Move your tongue and throat to move the groupings/lines of data up and down the graph. These are vocal size/resonance exercises you can practice with sight.

I got some free audio analysis software here. All I have is a webcam mic, but I'm pretty sure I see something on the spectrogram at 60, 120, and 240 hz. But I'm not sure if I'm going down the wrong path here.

If what you want is a real spectrum analyzer you'll need to try other software. For something free and open-source like Audacity, try Friture.

Your main focus should be work on balancing your vocal weight and vocal size and there are no good applications to help with that (there are spectrograms, but it's a rabbit hole - they are very hard to interpret and as practice shows, people do not benefit from spending time on trying to understand how to use them.) Above that, one of the first goals should be for you to learn how to hear the changes in size and weight directly, through auditory feedback, without any assistance, so using tools may even be detrimental (if you teach yourself to focus on them instead on honing your listening skills, that is.)