Description

I honestly had low expectations due that the vocals had a lot of echo and reverb. But hey, this model sounds good! this will probably be taken down for poor quality, but you can still find it on weights.gg if it gets taken down. Trained on 27 seconds of dataset ripped from YT. VA: Julius Scizzor (?) A transpose of -12 is recommended, but you can adjust to which sounds good in your opinion. Pitch extraction: RMVPE Steps: 2.55k Batch size: 2 Pretrain: Original v2 / 16k (due to the small cut-off frequency of the vocals in the dataset, i didn't expect that low cut-off....) Don't forget to credit my YT channel if used: @fitzyrvc2024; i would appreciate it. Huggingface: Weights.gg: