supple cedar
KLM 4.3 x3 Test Model
Aespa Karina Voice
"The index file is trained in Korean, so when using English or singing, make sure to set the specific index value to 0."
Dataset -
12 Mins of Speech
1 Min of Sing
Training Info -
Batch Size : 4
Total Epoch : 500
Speaker Channel :
s ID 0 : Speech Voice
s ID 1 : Vocal / Sing Voice
Model Link - https://huggingface.co/SeoulStreamingStation/RVC_Voice_Models/resolve/main/Karina_KLM43_x3_V5.zip?download=true
rapid eagle