{
  "file": "references/0055.wav",
  "sample_rate_for_analysis": 16000,
  "duration_sec": 4.122,
  "voiced_fraction": 0.789,
  "pause_map": [
    {
      "start": 2.21,
      "end": 2.71,
      "duration": 0.5
    },
    {
      "start": 3.17,
      "end": 3.69,
      "duration": 0.52
    }
  ],
  "f0_hz": {
    "median": 180.169,
    "p10_p90": [
      65.566,
      216.748
    ],
    "contour_20_points": [
      128.51,
      184.91,
      222.18,
      185.98,
      171.33,
      188.71,
      185.98,
      185.04,
      198.36,
      201.11,
      146.71,
      65.0,
      88.35,
      169.51,
      155.49,
      73.41,
      68.87,
      191.6,
      155.34,
      162.85
    ]
  },
  "energy_db": {
    "p10_p50_p90": [
      -66.601,
      -15.688,
      -9.394
    ],
    "dynamic_range_db": 57.208,
    "contour_20_points": [
      -70.07,
      -12.83,
      -12.82,
      -10.71,
      -12.75,
      -13.67,
      -10.4,
      -10.06,
      -25.91,
      -9.27,
      -25.53,
      -62.36,
      -66.6,
      -9.03,
      -13.35,
      -57.75,
      -66.62,
      -34.74,
      -12.47,
      -32.18
    ]
  },
  "timbral_proxies": {
    "spectral_centroid_hz_median": 1719.65,
    "spectral_flatness_median": 0.004
  },
  "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": "Аварийный люк откроется через три... Два... Один.",
  "text_chars_per_sec": 11.886,
  "source": {
    "file": "sources_silero/0055.wav",
    "sample_rate_for_analysis": 16000,
    "duration_sec": 3.275,
    "voiced_fraction": 0.665,
    "pause_map": [
      {
        "start": 1.74,
        "end": 1.9,
        "duration": 0.16
      },
      {
        "start": 2.27,
        "end": 2.64,
        "duration": 0.37
      }
    ],
    "f0_hz": {
      "median": 201.649,
      "p10_p90": [
        172.232,
        236.092
      ],
      "contour_20_points": [
        207.56,
        206.36,
        272.85,
        218.1,
        229.37,
        232.51,
        223.74,
        210.77,
        186.1,
        175.94,
        173.64,
        210.58,
        182.79,
        177.83,
        193.61,
        209.33,
        193.89,
        160.02,
        181.74,
        181.74
      ]
    },
    "energy_db": {
      "p10_p50_p90": [
        -76.527,
        -24.111,
        -13.957
      ],
      "dynamic_range_db": 62.57,
      "contour_20_points": [
        -68.85,
        -13.95,
        -14.3,
        -14.32,
        -29.03,
        -28.33,
        -13.86,
        -22.2,
        -21.79,
        -38.12,
        -38.14,
        -50.63,
        -23.53,
        -31.24,
        -78.78,
        -84.76,
        -21.95,
        -24.27,
        -60.92,
        -79.7
      ]
    },
    "timbral_proxies": {
      "spectral_centroid_hz_median": 1322.038,
      "spectral_flatness_median": 0.004
    },
    "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": "Аварийный люк откроется через три... Два... Один.",
    "text_chars_per_sec": 14.962
  },
  "pipeline_control": {
    "tts_text": "Аварийный люк откроется через три... Два... Один.",
    "duration_stage": "pre",
    "speaker_reference": "0055"
  }
}