Skip to content

Pretrained Models

Jump to bottom
Yist edited this page Dec 16, 2020 · 1 revision

The following models were trained on VoxCeleb1 dataset.

Model details:

  • 40-dim mel spectrogram as input
  • 3 layers of LSTM with hidden dimensions being 256
  • 256-dim speaker embedding

Download links:

Clone this wiki locally