Description

Link - I know there already is a model of this voice but its not as good as this one. The dataset is only 16 seconds long, so I had to make sure it's as clean as possible (voice lines were ripped from game files). First, I got rid of noise, then I cut out the imperfections. Next, I resampled the voice lines to 1200 Hz because anything above that is interpolation. Then, I applied new interpolation (I did that to improve audio quality because the original files were heavily compressed). Additionally, I ripped the voice lines of DogDay from the cartoon, isolated it again, and applied interpolation so that it matches with the cutout voice lines. I also did a little bit of EQ to match it as closely as possible to the cutout voice lines. Sample Rate 20k Also, as a side note, don't use the index file; it messes up the voice.