Independent-Study

Implement CycleGAN for singers voice conversion

Cycle GAN Structure

Web Application

The application based on the model can be found here

Task List

• Extract vocal from the music
• Remove slience part,segment and reformat the music
• Training the model using original CycleGAN model
• Create a demo web application
• Noise reduction of the converted sound segment using ML
• Improve the performance of the CycleGAN using Res2Net block
• Try many-to-one sound conversions with other models