Your voice memo, with a band under it.
A voice memo from your phone, or sing one now. Humming works — it only needs to hear the melody.
Found by counting, not guessing. Every one of these is editable — if it read you wrong, say so and the rest follows.
Tuning snaps to your scale, not the nearest chromatic note — that is the difference between corrected and robotic. Anything more than a tone off is left alone, because at that point it is a different note.
Rendered offline — a real file, not a recording of playback. Stems too, so it is a starting point rather than a dead end.