By the time we first learn to speak, we have already heard and stored the phonemes in our environment in our speech centre. This early-life programming takes place up to the age of approximately 18 months, after which the process in complete.
Phonemes learned after this age are stored in a different memory location (to be very brief about a very complex subject) and form part of our language, rather than our speech, system.
The trouble is that if we are singing lyrics in a language which is native to us, we have to have a very robust 'diversion' system in place which prevents the brain from simply reverting to the speech-centre (easier!) and which keeps feeding the 'singing vowels' into the system as required (more difficult for the brain).
This 'diversion' system cannot be mentally based as any attempt by the singer to 'make' the voice do somethinig by direct control is doomed to immediate failure. (q.v. my comments about thrying to control the focus of the eyes). Rather, we need to set up a physical, habit-based system which 'breaks' the body's normal speech habit and facilitates rather the production of the sounds required for singing.