Goals for now

train this small model on all text possible data (from pdf to text files)

when it works :

take the voice libraries and link them to this model

we can take it like the main model and start experimenting