Closed aedocw closed 7 months ago
I'm not sure if this could have anything to do with it, although one thing I've noticed while testing the problematic branch is that the progress spoken so far text percentage is significantly reduced when using vits (I assume this is a side affect of removing silence). Not sure if this is having an affect on xtts, although this seems to work fine otherwise.
This was resolved with the "more-bad-timing" branch. I was accidentally using the same index in two nested for loops.
This did not happen when I did the same thing without the xtts options.