I think I found a better and more natural solution than the original dialogue idea.
As far as that FakeYou site. It can produce decent results, but i find the GitHub release much better and more powerful. Sometimes the voices can be rendered almost robotic through the site, requiring a lot of post-render work in Audacity and Vegas.
However, audio trickery - and especially dialogue - is always difficult. It took me nearly two hours just to produce this snippet - and I don’t know if it will work, but here it is: