I have recently been working on creating a model using a large dataset. Using a basic voice tone, I recorded a 40 mins script and then applied dubbing over it. I did this dubbing three times, including variations with distinct voice changes such as [normal], [sad], [shouting], and [falsetto]. Each variation was recorded for the same duration, and the dataset was labeled accordingly, resulting in a dataset composed of a total of 20,000 chunks. Additionally, I used the same method to record songs, applying dubbing over them, but unlike the speech data, I used different singing techniques such as [falsetto] and [chest voice]. Initially, this model was intended for sale at the master shop, but as there were many people curious about the output of a model created with a large dataset, I decided to release it for sharing. Note that this model does not have a separate feature index file.

Train info :
RVC 2 / RVMPE
170 Epochs
Batch size per GPU : 6
Total Steps : N/A

Data set :
Total 170+ Mins of data

Model Link - https://huggingface.co/SeoulStreamingStation/RVC_Voice_Models/resolve/main/Voice_Nell_V2.zip?download=true

Additional Details:

Tags: No tags available

Download Link: https://huggingface.co/SeoulStreamingStation/RVC_Voice_Models/resolve/main/Voice_Nell_V2.zip?download=true