***"Live from Starr Park!"***

Voice of character is intentional, please don't take down for poor quality!

I realized that Mangio-Crepe doesn't just work on soft voices and short datasets, it also works amazingly with robotic voices. I now learned that Mangio-Crepe is actually underrated and works very good with short datasets and any type of voices (?).

Dataset contains 1 minute and 7 seconds of clean dialogue ripped from R-T's voicelines. Training involved 420 epochs and 9660 steps using Mangio-Crepe as pitch extraction algorithm at 32k sample rate with a batch size of 2 and the TITAN-Medium pretrain.

*I recommend changing the transpose from 0 to 12, but you can adjust to which sounds good in your opinion. Also don't rise up too high the index ratio just in case (keep it from 0.5 to 0.75), and for inference use RMVPE.*

Trained on RVC Disconnected and locally inferenced on Applio since that Mainline doesn't have Mangio-Crepe.

Don't forget to credit my YT channel if used: @Mantrax-IA; i would appreciate it.

Huggingface: https://huggingface.co/FurnTheFurnace/RVCModels/resolve/main/R-T-Brawl-Stars-Mangio_e420_s9660.zip?download=true
Weights.gg: https://www.weights.gg/models/clyjglzd50phc13xcuazpsgjz

Additional Details:

Tags: No tags available

Download Link: https://huggingface.co/FurnTheFurnace/RVCModels/resolve/main/R-T-Brawl-Stars-Mangio_e420_s9660.zip?download=true