{
    "sample_rate":  16000,
    "latent_hop":  320,
    "chunk_ms":  2400,
    "current_ms":  320,
    "future_ms":  100,
    "smooth_ms":  20,
    "ref_seconds":  8.0,
    "frame_mel":  {
                      "n_fft":  1024,
                      "win_length":  640,
                      "hop_length":  320,
                      "mel_fmin":  10,
                      "mel_fmax":  null,
                      "num_mels":  128,
                      "power":  1,
                      "norm":  "slaney",
                      "mel_scale":  "slaney",
                      "log":  "amplitude_to_db",
                      "top_db":  80.0,
                      "eps":  1E-09
                  },
    "speaker_fbank":  {
                          "num_mels":  80,
                          "sample_rate":  16000,
                          "mean_nor":  true
                      },
    "volume_normalize":  true,
    "highpass_cutoff_freq":  40,
    "whisper_mel":  {
                        "n_fft":  400,
                        "hop_length":  160,
                        "feature_size":  128,
                        "sampling_rate":  16000,
                        "padding_value":  0.0,
                        "pad_to_multiple_of_samples":  320
                    },
    "shapes":  {
                   "t_lat_chunk":  120,
                   "t_tok_chunk":  30,
                   "t_out_chunk":  120,
                   "t_ref_mel":  401
               }
}