As far as making the shapes, I'd suggest recording a video of yourself lip syncing the line so you have reference to go by. You can then create a plane in the viewport and drag your video onto it so you can have a frame by frame video right next to the model. (Just make sure your lip sync matches the audio

) You can also watch the show that the clip is from but there isn't always convenient camera angles doing it that way. Making the shapes from scratch works fine of course but you tend to get a little more feeling into it when you are using reference.
edit: curses, he beat me!