Why not something akin to "fixed-offset cursor"? Actual cursor/pointer is always a centimeter above the finger position (thus always visible [bottom of the screen to be handled as a special case]), and finger movements manipulate position, while hold duration accesses alternate modes.
For a somewhat similar implementation (touchscreen is essentially handled as touchpad) look at how Teamviewer handles remote sessions to Windows desktop from smartphones.
I would really like any/all of those as a toggle-on mode for use on Android itself... selecting text with arthritis fingers is a pain – literally and figuratively.
Try out the app Quick Cursor[1] it does a lot of what you describe. Pretty intuitive too, you build muscle memory for it quickly.
1: https://play.google.com/store/apps/details?id=com.quickcurso...
The shape with sharp, jagged edges would occur in real life if it were made of some hard material, perhaps, like glass, metal, etc. The shape with the soft, curvy lines would occur if it were made of something softer and possibly elastic.
It doesn't take much intuition to guess that the sharper shape will produce a sound closer to "ki," with sharp transients and lots of high frequencies - like a piece of glass or metal falling - and the rounded one perhaps closer to "bou," with softer transients and perhaps a time-varying resonant frequency as the shape is malleable (think of the sound of a drop of water landing).