vocoder/config.json

{
  "_class_name": "LTX2VocoderWithBWEModel",
  "_diffusers_version": "0.37.0",
  "activation": "snakebeta",
  "apply_final_activation": true,
  "bwe_activation": "snakebeta",
  "bwe_apply_final_activation": false,
  "bwe_hop_length": 80,
  "bwe_input_sampling_rate": 16000,
  "bwe_n_fft": 512,
  "bwe_num_mel_channels": 64,
  "bwe_resblock": "AMP1",
  "bwe_resblock_dilation_sizes": [
    [
      1,
      3,
      5
    ],
    [
      1,
      3,
      5
    ],
    [
      1,
      3,
      5
    ]
  ],
  "bwe_resblock_kernel_sizes": [
    3,
    7,
    11
  ],
  "bwe_upsample_initial_channel": 512,
  "bwe_upsample_kernel_sizes": [
    12,
    11,
    4,
    4,
    4
  ],
  "bwe_upsample_rates": [
    6,
    5,
    2,
    2,
    2
  ],
  "bwe_use_bias_at_final": false,
  "bwe_use_tanh_at_final": false,
  "output_sampling_rate": 48000,
  "resblock": "AMP1",
  "resblock_dilation_sizes": [
    [
      1,
      3,
      5
    ],
    [
      1,
      3,
      5
    ],
    [
      1,
      3,
      5
    ]
  ],
  "resblock_kernel_sizes": [
    3,
    7,
    11
  ],
  "upsample_initial_channel": 1536,
  "upsample_kernel_sizes": [
    11,
    4,
    4,
    4,
    4,
    4
  ],
  "upsample_rates": [
    5,
    2,
    2,
    2,
    2,
    2
  ],
  "use_bias_at_final": false,
  "use_tanh_at_final": false
}