First of all, it's impressive, especially considering it's still in development. If you don't stop developing it too early and train the AI better it can become the best software of that kind.
So far I've only noticed one thing worth criticizing and it's that the censorship does not work on fullscreen in windows media player. The moment I turn on fullscreen, software stops censoring even though you can still hear the audio, so it certainly still works in some way.
I also feel that you could add a creation of presets so users would not have to delete or change everything when they want to try out something different.
Also, I think it's missing the game mode featured in your previous work. Maybe it didn't have many levels and became kinda underwhelming once you've reached the 3rd stage, but it was still a very fun feature. From what I read you plan on adding something similar so I'm just letting you know it was pretty cool.
Best of luck, can't wait to see how things will look in the future!