OK. It helps if you use the desktop version as opposed to the online one. This is only available for PC, which was a pain to me as I'm a Mac guy but I do have a PC for games so I used that.
This 'songwriter' video was just the straight text to speech function.
However, on the 'Sad About Us' video I started out by using the text to speech and then trying to synch the results to the audio in iMovie. That was a complete failure as the characters talk really fast in Xtranormal and singing is slower than normal speech. Even slowing the clips down made for poor results. I started again.
Because the video options are very limited in XN (no moving cameras etc.) I story-boarded the whole thing and shot each scene as a separate project, this was so I was later able to cut, pan, zoom etc using the facilities in iMovie. I then got the multi-track audio of the song and ran off all the vocal lines as separate .wav files which I loaded separately into each scene using the the 'voice over' facilty. This is when you load in the audio, and you also type in the text, then the program will attempt to match the two up. This was more work than it would be if it wasn't a duet of course, as I could only do one couplet at a time.
I had to convert the format of all the finished clips, as iMovie doesn't recognise the native format of Xtranormal: I used AVC for this (and it's a free program). Then I loaded the song wav itself into iMovie and then the clips as though they were real life footage. I manually lip-synced the audio to the pictures as I edited. It took a very long time all in all, but I am pleased with it - even though a few bits are dodgy. There's nothing you can do about that though, as you can't change which typed consonants the program chooses to align with the audio - and singing is different to the speech it expects.
Sorry if that reads like a bunch of gobbledygook but you will see what I mean if you try and do it. Please PM me if I can help.