The paper says that all audios are resampled at 16khz and it is trained on librilight. But the vae is working on 24khz latent?
The paper says that all audios are resampled at 16khz and it is trained on librilight. But the vae is working on 24khz latent?