{
  "file": "references/0349.wav",
  "sample_rate_for_analysis": 16000,
  "duration_sec": 2.011,
  "voiced_fraction": 0.634,
  "pause_map": [],
  "f0_hz": {
    "median": 143.416,
    "p10_p90": [
      65.0,
      164.739
    ],
    "contour_20_points": [
      167.62,
      167.62,
      164.57,
      164.74,
      140.67,
      149.33,
      147.17,
      144.67,
      138.53,
      148.26,
      161.91,
      164.74,
      155.26,
      138.23,
      135.04,
      109.08,
      78.38,
      65.03,
      65.0,
      65.0
    ]
  },
  "energy_db": {
    "p10_p50_p90": [
      -55.024,
      -14.614,
      -6.991
    ],
    "dynamic_range_db": 48.033,
    "contour_20_points": [
      -72.31,
      -58.45,
      -12.16,
      -7.26,
      -9.3,
      -7.77,
      -10.6,
      -15.88,
      -10.42,
      -15.0,
      -9.69,
      -6.37,
      -8.16,
      -12.09,
      -27.93,
      -14.19,
      -35.54,
      -49.23,
      -55.16,
      -60.32
    ]
  },
  "timbral_proxies": {
    "spectral_centroid_hz_median": 1689.533,
    "spectral_flatness_median": 0.008
  },
  "interpretation": {
    "target": "soft, close-mic, restrained mechanical delivery",
    "sarcasm": "low-to-moderate pitch movement with deliberate clause-final timing",
    "smile": "look for local energy/brightness lift around emphasized words; do not synthesize a broad cheerful smile"
  },
  "text": "Я так сильно вас ненавижу.",
  "text_chars_per_sec": 12.926,
  "source": {
    "file": "sources_silero/0349.wav",
    "sample_rate_for_analysis": 16000,
    "duration_sec": 1.675,
    "voiced_fraction": 0.75,
    "pause_map": [],
    "f0_hz": {
      "median": 213.639,
      "p10_p90": [
        172.53,
        257.014
      ],
      "contour_20_points": [
        258.5,
        258.5,
        243.32,
        231.04,
        246.83,
        248.03,
        251.86,
        258.29,
        257.8,
        243.51,
        218.04,
        210.0,
        231.31,
        214.62,
        194.31,
        185.23,
        168.59,
        172.19,
        172.53,
        180.69
      ]
    },
    "energy_db": {
      "p10_p50_p90": [
        -54.132,
        -22.61,
        -14.893
      ],
      "dynamic_range_db": 39.239,
      "contour_20_points": [
        -79.0,
        -65.0,
        -47.82,
        -14.96,
        -37.7,
        -16.98,
        -24.56,
        -21.99,
        -16.98,
        -13.53,
        -20.29,
        -15.42,
        -35.15,
        -20.98,
        -20.93,
        -22.92,
        -22.11,
        -28.64,
        -29.19,
        -53.43
      ]
    },
    "timbral_proxies": {
      "spectral_centroid_hz_median": 1518.037,
      "spectral_flatness_median": 0.006
    },
    "interpretation": {
      "target": "soft, close-mic, restrained mechanical delivery",
      "sarcasm": "low-to-moderate pitch movement with deliberate clause-final timing",
      "smile": "look for local energy/brightness lift around emphasized words; do not synthesize a broad cheerful smile"
    },
    "text": "Я так сильно вас ненавижу.",
    "text_chars_per_sec": 15.522
  },
  "pipeline_control": {
    "tts_text": "Я так сильно вас ненавижу.",
    "duration_stage": "pre",
    "speaker_reference": "0349"
  }
}