In this post, we show the demo of EAD-VC.

Model Architecture

Figure.1 The architecture of the EAD-VC.

Experiment

Results

In order to evaluate and show the performance of our model, we list the four Tables to respective show the four Female-to-Male conversion (F2M), Male-to-Female conversion (M2F), One-Shot voice conversion (Timbre Conversion and Not Only Timbre Conversion).

 

Table.1 The comparision between the EAD-VC and other models on F2M.

ID	Ground Truth	Adain-VC	SpeechFlow	Liu et al.	VQMIVC	EAD-VC
Sample 1
Sample 2
Sample 3

 

Table.2 The comparision between the EAD-VC and other models on M2F.

ID	Ground Truth	Adain-VC	SpeechFlow	Liu et al.	VQMIVC	EAD-VC
Sample 1
Sample 2
Sample 3

 

Table.3 The comparision on One-Shot voice conversion.

Text	Reference Timbre	Adain-VC	SpeechFlow	Liu et al.	VQMIVC	EAD-VC
That's no bad thing.
I love the lab of Large Audio Model.

 

Table.4 The comparision on One-Shot voice conversion (Not Only Timbre Conversion)

Type	Pitch	Rhythm	Timbre	Pitch+Rhythm	Pitch+Timbre	Rhythm+Timbre	Pitch+Rhythm+Timbre
Liu et al.
EAD-VC