problem seperating instrumental

[Hulababa-22]

so i have a problem with seperating the instrument from the vocal when i go to song output and hear the instrument only sound it still have the vocal from the song in there so it doesn't seperate the vocal good enough but when i tried to seperate it myself manually using UVR and using the same model it does a very good job in seperating it so why does it do a terrible job in aicovergen(im running locally)? maybe you can make it use another vocal seperation model(like kim vocal 2) or add an option to change the model it uses, and settings for the seperation or somthing?