word-sub / tests /expected /corner_cases /large-v2.accurate_gloria.mp3.words.json
RaeesRSB's picture
Upload 185 files
2cba4ca
{
"text": " Ella, my glorious love, how are you? Oh, I'm okay. I will be. I said she could stay with us tomorrow just until she feels better. Of course she can. No, this won't be for long. Well, if you can stay as long as you want, my love. I've really missed you. Pops.",
"segments": [
{
"id": 0,
"seek": 0,
"start": 0.0,
"end": 6.28,
"text": " Ella, my glorious love, how are you?",
"tokens": [
50364,
29261,
11,
452,
24026,
959,
11,
577,
366,
291,
30,
50664
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.452,
"words": [
{
"text": "Ella,",
"start": 0.0,
"end": 1.68,
"confidence": 0.126
},
{
"text": "my",
"start": 1.94,
"end": 2.7,
"confidence": 0.368
},
{
"text": "glorious",
"start": 2.7,
"end": 3.5,
"confidence": 0.729
},
{
"text": "love,",
"start": 3.5,
"end": 4.18,
"confidence": 0.432
},
{
"text": "how",
"start": 4.94,
"end": 5.66,
"confidence": 0.693
},
{
"text": "are",
"start": 5.66,
"end": 6.02,
"confidence": 0.902
},
{
"text": "you?",
"start": 6.02,
"end": 6.28,
"confidence": 0.845
}
]
},
{
"id": 1,
"seek": 0,
"start": 6.28,
"end": 9.18,
"text": " Oh, I'm okay. I will be.",
"tokens": [
50664,
876,
11,
286,
478,
1392,
13,
286,
486,
312,
13,
50814
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.68,
"words": [
{
"text": "Oh,",
"start": 6.28,
"end": 6.6,
"confidence": 0.427
},
{
"text": "I'm",
"start": 6.68,
"end": 7.02,
"confidence": 0.958
},
{
"text": "okay.",
"start": 7.02,
"end": 7.64,
"confidence": 0.605
},
{
"text": "I",
"start": 8.3,
"end": 8.58,
"confidence": 0.562
},
{
"text": "will",
"start": 8.58,
"end": 8.86,
"confidence": 0.813
},
{
"text": "be.",
"start": 8.86,
"end": 9.18,
"confidence": 0.871
}
]
},
{
"id": 2,
"seek": 0,
"start": 9.18,
"end": 11.48,
"text": " I said she could stay with us tomorrow just until she feels better.",
"tokens": [
50814,
286,
848,
750,
727,
1754,
365,
505,
4153,
445,
1826,
750,
3417,
1101,
13,
50964
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.571,
"words": [
{
"text": "I",
"start": 9.18,
"end": 9.44,
"confidence": 0.66
},
{
"text": "said",
"start": 9.44,
"end": 9.62,
"confidence": 0.869
},
{
"text": "she",
"start": 9.62,
"end": 9.76,
"confidence": 0.748
},
{
"text": "could",
"start": 9.76,
"end": 9.88,
"confidence": 0.813
},
{
"text": "stay",
"start": 9.88,
"end": 10.08,
"confidence": 0.925
},
{
"text": "with",
"start": 10.08,
"end": 10.22,
"confidence": 0.808
},
{
"text": "us",
"start": 10.22,
"end": 10.34,
"confidence": 0.628
},
{
"text": "tomorrow",
"start": 10.34,
"end": 10.56,
"confidence": 0.219
},
{
"text": "just",
"start": 10.56,
"end": 10.74,
"confidence": 0.067
},
{
"text": "until",
"start": 10.74,
"end": 10.86,
"confidence": 0.4
},
{
"text": "she",
"start": 10.86,
"end": 10.98,
"confidence": 0.803
},
{
"text": "feels",
"start": 10.98,
"end": 11.18,
"confidence": 0.788
},
{
"text": "better.",
"start": 11.18,
"end": 11.48,
"confidence": 0.807
}
]
},
{
"id": 3,
"seek": 0,
"start": 11.5,
"end": 13.34,
"text": " Of course she can.",
"tokens": [
50964,
2720,
1164,
750,
393,
13,
51014
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.333,
"words": [
{
"text": "Of",
"start": 11.5,
"end": 12.1,
"confidence": 0.123
},
{
"text": "course",
"start": 12.1,
"end": 12.6,
"confidence": 0.86
},
{
"text": "she",
"start": 12.6,
"end": 12.88,
"confidence": 0.085
},
{
"text": "can.",
"start": 12.88,
"end": 13.34,
"confidence": 0.674
}
]
},
{
"id": 4,
"seek": 0,
"start": 13.34,
"end": 15.2,
"text": " No, this won't be for long.",
"tokens": [
51014,
883,
11,
341,
1582,
380,
312,
337,
938,
13,
51114
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.649,
"words": [
{
"text": "No,",
"start": 13.34,
"end": 13.78,
"confidence": 0.328
},
{
"text": "this",
"start": 13.9,
"end": 14.24,
"confidence": 0.419
},
{
"text": "won't",
"start": 14.24,
"end": 14.54,
"confidence": 0.917
},
{
"text": "be",
"start": 14.54,
"end": 14.68,
"confidence": 0.906
},
{
"text": "for",
"start": 14.68,
"end": 14.88,
"confidence": 0.884
},
{
"text": "long.",
"start": 14.88,
"end": 15.2,
"confidence": 0.823
}
]
},
{
"id": 5,
"seek": 0,
"start": 15.2,
"end": 17.44,
"text": " Well, if you can stay as long as you want, my love.",
"tokens": [
51114,
1042,
11,
498,
291,
393,
1754,
382,
938,
382,
291,
528,
11,
452,
959,
13,
51214
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.69,
"words": [
{
"text": "Well,",
"start": 15.2,
"end": 15.46,
"confidence": 0.553
},
{
"text": "if",
"start": 15.52,
"end": 15.6,
"confidence": 0.084
},
{
"text": "you",
"start": 15.6,
"end": 15.68,
"confidence": 0.898
},
{
"text": "can",
"start": 15.68,
"end": 15.8,
"confidence": 0.827
},
{
"text": "stay",
"start": 15.8,
"end": 16.04,
"confidence": 0.922
},
{
"text": "as",
"start": 16.04,
"end": 16.18,
"confidence": 0.896
},
{
"text": "long",
"start": 16.18,
"end": 16.32,
"confidence": 0.893
},
{
"text": "as",
"start": 16.32,
"end": 16.46,
"confidence": 0.902
},
{
"text": "you",
"start": 16.46,
"end": 16.58,
"confidence": 0.905
},
{
"text": "want,",
"start": 16.58,
"end": 16.76,
"confidence": 0.731
},
{
"text": "my",
"start": 16.88,
"end": 17.02,
"confidence": 0.901
},
{
"text": "love.",
"start": 17.02,
"end": 17.44,
"confidence": 0.831
}
]
},
{
"id": 6,
"seek": 0,
"start": 17.44,
"end": 18.96,
"text": " I've really missed you.",
"tokens": [
51214,
286,
600,
534,
6721,
291,
13,
51314
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.529,
"words": [
{
"text": "I've",
"start": 17.44,
"end": 17.58,
"confidence": 0.309
},
{
"text": "really",
"start": 17.58,
"end": 18.18,
"confidence": 0.763
},
{
"text": "missed",
"start": 18.18,
"end": 18.68,
"confidence": 0.778
},
{
"text": "you.",
"start": 18.68,
"end": 18.96,
"confidence": 0.622
}
]
},
{
"id": 7,
"seek": 0,
"start": 19.58,
"end": 20.12,
"text": " Pops.",
"tokens": [
51314,
430,
3370,
13,
51364
],
"temperature": 0.0,
"avg_logprob": -0.3579153639546941,
"compression_ratio": 1.425414364640884,
"no_speech_prob": 0.24429768323898315,
"confidence": 0.02,
"words": [
{
"text": "Pops.",
"start": 19.58,
"end": 20.12,
"confidence": 0.02
}
]
}
],
"language": "en"
}