Question about your BSRoformer models.
Hey there, I have a question. If you do plan on training a new BSRoformer or MelBandRoformer model sometime in the future, can you train a model on Friday Night Funkin' songs? I'm trying to get rid of some distorted FNF vocals from a song I ran inference on using different models that you created in the past.
What I meant by distorted FNF vocals is some Friday Night Funkin' mods have songs that have distorted vocals. I think you can find some fnf songs with separate instrumental and vocals stems on some websites or wikis as well.
I welcome suggestions for musics and vocals to include in the dataset.
However, since I won’t have access to my main computer for at least two months, it may be some time before I can train any models.
Okay, I understand. When you do get your computer back, do you know what instrumentals and vocals you used in your dataset for your MelbandRoformer V2 you trained? If you have it, do you mind dropping the dataset if possible?
I do not intend to make the dataset publicly available.
This is because it contains datasets shared with me by others as well as data I collected myself.