{
  "model": "tiny.en",
  "device": "cpu",
  "compute_type": "int8",
  "word_timestamps": false,
  "model_load_seconds": 19.622,
  "inference_seconds": 1.444,
  "duration": 9.472125,
  "language": "en",
  "segments": [
    {
      "id": 1,
      "seek": 0,
      "start": 0.0,
      "end": 2.2,
      "text": " This is a local tutorial test.",
      "tokens": [
        50363,
        770,
        318,
        257,
        1957,
        11808,
        1332,
        13,
        50473
      ],
      "avg_logprob": -0.22426387236306541,
      "compression_ratio": 1.236842105263158,
      "no_speech_prob": 0.04013679176568985,
      "words": null,
      "temperature": 0.0
    },
    {
      "id": 2,
      "seek": 0,
      "start": 2.2,
      "end": 7.16,
      "text": " We render a short video, check its frame rate, and export word timestamps.",
      "tokens": [
        50473,
        775,
        8543,
        257,
        1790,
        2008,
        11,
        2198,
        663,
        5739,
        2494,
        11,
        290,
        10784,
        1573,
        4628,
        395,
        9430,
        13,
        50721
      ],
      "avg_logprob": -0.22426387236306541,
      "compression_ratio": 1.236842105263158,
      "no_speech_prob": 0.04013679176568985,
      "words": null,
      "temperature": 0.0
    },
    {
      "id": 3,
      "seek": 0,
      "start": 7.16,
      "end": 9.24,
      "text": " The original audio stays unchanged.",
      "tokens": [
        50721,
        383,
        2656,
        6597,
        14768,
        21588,
        13,
        50825
      ],
      "avg_logprob": -0.22426387236306541,
      "compression_ratio": 1.236842105263158,
      "no_speech_prob": 0.04013679176568985,
      "words": null,
      "temperature": 0.0
    }
  ]
}