Contains models extracted from Ace Combat 5 main character voicelines. The trained audio was not filtered for noise, any inference from the models will have in-game "radio static" and other noise.
Original dataset per voice is < 10mins including noise, so results are not high quality.