Normally you would take the UTAU program, draw the notes with the lyrics and start playing with them up and down along with the portamento so it resembles human speech. It's quite complex and you need good hearing for that so it doesn't sound like a robot. I myself feel not capable of doing such a thing. Lately I have seen there is a new tendency: VCVC. Some people tried to play with UTAU samples to make it more like talkloid, but I do not have enough information to tell you the depths of this new system.