48k is a bit sweet, supports singing AI girl Liuli RVC model

Note: This model may have some background noise; please use it together with the NS1 effect (output) on the rack.

The advantage of this model is that it sings more naturally and is great for chatting, but not for a bubble voice.

The listening effect on the right is for real-time inference, with an added ns1 in the output channel to suppress background noise.

If you mind, you can also click on my homepage to view other noise-free models.