uncut forge
First of all, I apologize for the thumbnail; I showed it to my nephew, and he got scared and cried. Anyway, it all started when two datasets accidentally went into the same folder during the saving process, creating a model by mistake. But I was surprised to see it sounding quite natural.
So, I mixed a total of 8 voice models and put them together, and this was the result.
Well, it's just a model made for fun, but it seems it could save some time for our catfish(넷마카) friends living quietly around the world, constantly changing models. Simply adjusting the pitch can change the tone of the voice, after all.
Information :
Model :
Model Name : Mixture of Human Voice
Train info :
RVC 2 / RVMPE
550 Epochs
Batch size per GPU : 38
Dataset - source Info :
14 Hours of Speech
2 Hours of Sing
Dataset - source correction :
Pretrained by KLM v7
Model Link - https://huggingface.co/SeoulStreamingStation/IU-Voice-MultiLanguage-V1/resolve/main/MixtureofHuman.zip?download=true
quick ginkgo
patent marsh
fringe solstice
little condor
sturdy estuary