{
  "feature_extractor_type": "VocosFeatureExtractor",
  "feature_size": 100,
  "hop_length": 256,
  "n_fft": 1024,
  "num_mel_bins": 100,
  "padding": "center",
  "padding_side": "right",
  "padding_value": 0.0,
  "return_attention_mask": false,
  "sampling_rate": 24000
}