Description

40k sample rate, batch size 8, 8 minute dataset, trained with RMVPE. I made this because a meme in the italian community consists of random people joining and asking for a model of Rondodasosa, which led me to make one just to make the requests stop, but it's not one of my best models. Trained on his screaming vocals, you might need to raise the pitch during inferences to get the vocals right. Per gli italiani: urla molto e il modello basato su quello, se fate le cover vi consiglio di provare ad alzare il pitch finch il risultato non vi soddisfa. Enjoy!