pipeline_config.json

{
  "model": "kokoro-82m",
  "version": "v2",
  "stages": [
    "duration",
    "prosody",
    "decoder"
  ],
  "max_phonemes": 128,
  "phoneme_buckets": [
    16,
    32,
    64,
    128
  ],
  "d_transposed_dim": 640,
  "t_en_dim": 512,
  "style_dim": 128,
  "sample_rate": 24000,
  "samples_per_frame": 600,
  "buckets": {
    "5s": {
      "frames": 200,
      "samples": 120000
    },
    "10s": {
      "frames": 400,
      "samples": 240000
    },
    "15s": {
      "frames": 600,
      "samples": 360000
    }
  }
}