trained on 4 minutes and 42 seconds of data from the portal reloaded mod. (rx de-ess, mouth de-click, resample + eq, noise gate) - dataset provided by: @mystic sapphire
cv. Harry101UK
pitch extraction: mangio-crepe (applio crepe)
hop length: 8
steps: 5.04k
batch size: 6
pretrain: original v2 / 32k
precision: fp16
huggingface: https://huggingface.co/FurnTheFurnace/RVCModels/resolve/main/Announcer-PORTAL-Reloaded_630e_5040s.zip?download=true
model is kinda strange at singing, recommended use for speech. rmvpe and a index ratio of 0.5 - 0.75 are recommended to use for inference.
please don't forget to credit me when you use this model. please don't reupload any of my models without my permission, otherwise i'll report them to be taken down. for now on, my models are now licensed with a custom license. check it out in: https://huggingface.co/FurnTheFurnace/RVCModels/blob/main/LICENSE.

