[ a bit late, but I completely missed this thread ]

Originally Posted By: Planobilly
I am beginning to think that AI is in some ways similar to a human vocalist. One quickly reaches the limits of the vocal range and that requires adjusting the key signature.

This!!!

I've posted this somewhere before, but anyway: treat synthetic singers just like a real singer!
This means :
  • find her (assuming female vocals) usable vocal range
  • find her best sounding vocal range. If needed, transpose your song
  • find a singing rhythm that suits both their singing and the song
  • don't request things she can't deliver.

The question of how to start a song is as old as music itself, and everyone does it differently. What works best for me:
  • I noodle around on my acoustic or my keyboard until I find an interesting chord progression
  • Enter the chords into BIAB and play with some styles and instruments until it sounds good (or is discarded).
  • Decide on the structure of the song
  • Make a raw mix in my DAW
  • Just sing along until the lyrics come. Really. Unlike most, I can't start with a 'lalala' melody because I ALWAYS change the melody with the final lyrics.
    I don't know why I'm wired this way.
    I use the VST plugin of SynthV in my DAW and enter the lyrics and notes directly there.
  • I have absolutely no idea where my harmonies come from. I sing along and they are just there.
    Must be inherited from my mother who is a harmony monster smile
  • Mix until it sounds like a song, post it here, rinse and repeat

Like I said, this works for me, but not necessarily for you.
My nephew does it the other way around, he starts with the vocals and the melody and builds everything around that.
So, even in my own family, there's no agreement on how to write a song.

Last edited by B.D.Thomas; 02/05/23 03:24 AM.
Goldmania Bella
Listen on Spotify
Watch on YouTube
New on YouTube: Deeply Sad Music