license: openrail
Another model aside from Princess Bubblegum, I made a model for Flame Princess since she's a rapper since Season 10 on Adventure Time. So why not give her a shot to make one for her? I think she can sing Eminem's song if she can.
Trained her in 125 epochs, using RMVPE algorithm. Datasets: 14 minutes from her 13 major episode appearances.
Also, my tablet overheated so bad for finding samples for FP since she's a hot momma. What a coincidence hehehe.
Also, it's still certified pinoy-made model by KennyFromPH.
My Channel: https://youtube.com/@KennyFromPH07
I don't wanna make an epoch update anymore since I'm tired of waiting. Also my tablet's performance is now deteriorating for being awake for hours without restarting. I used UVR5 Colab Edition for vocal isolation this time, and yes I still trained the model in my old tablet. Also, it takes me 3 times since the 'train feature index' refuses to train the index file. So, I have to reprocess the dataset, then re-extract the pitch extraction all over to add the index desperately.