"Uma Musume" supports Chinese Mambo/poetic drama, singable, loli voice RVC model

Not long ago, I saw a model of a poetry and opera in Miaoyin. I downloaded it and tried it out, and found that besides some noise, it also doesn't support Chinese, and the tone is more girlish.

So during the holidays, I made a model of a poetry opera with a loli-like tone for everyone.

This model supports both Chinese and Japanese, and it leans toward the voice tone that's currently popular in short videos. I collected about 40 minutes of dataset and then inferred it into Chinese, so the "Da Zuo" sound won't appear when used.

Of course, I also created GPT models: "Uma Musume" Mambo/Poetic Opera GPT-SoVITS model

Finally, I will include a preview of the singing part.

Before reasoning

<!--[if lt IE 9]>document.createElement('audio');<![endif]-->

After reasoning

Singing part (Huangfengling male-to-female +12 tone change).

Parameters

Training log