Description:
The model was trained on 10 minutes of Twilight Sparkle lines and one minute of song.
To get the model email me on my discord - yuduzfridoed367
The model will most likely be improved. This is an intermediate version that can talk and sing quite well.
The vocals of the music in the video were obtained through a neural network. If you record pure audio on a microphone, the result will naturally be better.