I had a similar thought until I actually started learning an instrument and my SO, flutist graduated from an Italian conservatory, made me start singing with the right intonation (singing the same notes you are playing on the instrument) and rhythm (division and subdivision alongside a metronome). Long story short: singing is hard and requires real effort. There is a good chance that what you think you can articulate is probably far away from minimal quality on both pitch and time.

While that's true enough for many people, adding a feedback step and some practice would improve matters quickly.

There is also a good chance someone can learn to encode quite a bit into vocalizations.

"some practice would improve matters quickly"

Anyone who seriously learned an instrument or to sing, knows that it takes very hard practice for years to improve.

I have done that myself. Both vocal and instrument.

Yes, but vocalizations may well improve more quickly.

Frankly, I have toyed with this kind of thing. Software could do a lot, in particular, allow the user to adapt their expressions in various ways.

You can to some extent. There are a number of tools that attempt to convert from audio to note information (MIDI).

As someone who's done some experiments with those, there are some major caveats:

- Most people can't sing a melody as well as they think they can. Trying to figure out what they were trying to do is where most of the effort goes. Even I'm mediocre at such, and I'm at a professional level at one instrument, and competent at a few others.

- You have to deal with overtones. Almost all naturally produced notes are a stack of waveforms, and figuring out which one caries the intended signal isn't always trivial (and can be even more complicated in polyphonic harmony).