Audio-reactive Latent Interpolations with StyleGAN