whisper-timestamped / tests /expected /tiny_auto /gloria.mp3.words.json
kullup's picture
Upload 187 files
9480b39 verified
{
"text": " Hello. You're my girlfriend. How are you? I'm okay. Why will be? I said she could stay with us, but she feels better. Oh, she can. There's more to be for long. What if you can stay as long as you want? My girlfriend. Really, Missia? I agree.",
"segments": [
{
"id": 0,
"seek": 0,
"start": 1.34,
"end": 1.78,
"text": " Hello.",
"tokens": [
50364,
2425,
13,
50464
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.408,
"words": [
{
"text": "Hello.",
"start": 1.34,
"end": 1.78,
"confidence": 0.408
}
]
},
{
"id": 1,
"seek": 0,
"start": 1.96,
"end": 4.26,
"text": " You're my girlfriend.",
"tokens": [
50464,
509,
434,
452,
10369,
13,
50614
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.176,
"words": [
{
"text": "You're",
"start": 1.96,
"end": 3.12,
"confidence": 0.193
},
{
"text": "my",
"start": 3.12,
"end": 3.2,
"confidence": 0.206
},
{
"text": "girlfriend.",
"start": 3.2,
"end": 4.26,
"confidence": 0.124
}
]
},
{
"id": 2,
"seek": 0,
"start": 5.52,
"end": 6.54,
"text": " How are you?",
"tokens": [
50614,
1012,
366,
291,
30,
50714
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.627,
"words": [
{
"text": "How",
"start": 5.52,
"end": 5.8,
"confidence": 0.488
},
{
"text": "are",
"start": 5.8,
"end": 6.16,
"confidence": 0.525
},
{
"text": "you?",
"start": 6.16,
"end": 6.54,
"confidence": 0.96
}
]
},
{
"id": 3,
"seek": 0,
"start": 6.76,
"end": 7.8,
"text": " I'm okay.",
"tokens": [
50714,
286,
478,
1392,
13,
50764
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.836,
"words": [
{
"text": "I'm",
"start": 6.76,
"end": 7.1,
"confidence": 0.838
},
{
"text": "okay.",
"start": 7.1,
"end": 7.8,
"confidence": 0.83
}
]
},
{
"id": 4,
"seek": 0,
"start": 8.38,
"end": 9.14,
"text": " Why will be?",
"tokens": [
50764,
1545,
486,
312,
30,
50814
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.383,
"words": [
{
"text": "Why",
"start": 8.38,
"end": 8.66,
"confidence": 0.344
},
{
"text": "will",
"start": 8.66,
"end": 8.96,
"confidence": 0.522
},
{
"text": "be?",
"start": 8.96,
"end": 9.14,
"confidence": 0.313
}
]
},
{
"id": 5,
"seek": 0,
"start": 9.4,
"end": 11.52,
"text": " I said she could stay with us, but she feels better.",
"tokens": [
50814,
286,
848,
750,
727,
1754,
365,
505,
11,
457,
750,
3417,
1101,
13,
50964
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.692,
"words": [
{
"text": "I",
"start": 9.4,
"end": 9.54,
"confidence": 0.285
},
{
"text": "said",
"start": 9.54,
"end": 9.66,
"confidence": 0.621
},
{
"text": "she",
"start": 9.66,
"end": 9.82,
"confidence": 0.845
},
{
"text": "could",
"start": 9.82,
"end": 9.98,
"confidence": 0.911
},
{
"text": "stay",
"start": 9.98,
"end": 10.14,
"confidence": 0.992
},
{
"text": "with",
"start": 10.14,
"end": 10.26,
"confidence": 0.986
},
{
"text": "us,",
"start": 10.26,
"end": 10.44,
"confidence": 0.781
},
{
"text": "but",
"start": 10.64,
"end": 10.72,
"confidence": 0.228
},
{
"text": "she",
"start": 10.72,
"end": 10.94,
"confidence": 0.971
},
{
"text": "feels",
"start": 10.94,
"end": 11.24,
"confidence": 0.76
},
{
"text": "better.",
"start": 11.24,
"end": 11.52,
"confidence": 0.987
}
]
},
{
"id": 6,
"seek": 0,
"start": 12.18,
"end": 13.44,
"text": " Oh, she can.",
"tokens": [
50964,
876,
11,
750,
393,
13,
51064
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.286,
"words": [
{
"text": "Oh,",
"start": 12.18,
"end": 12.6,
"confidence": 0.18
},
{
"text": "she",
"start": 12.8,
"end": 12.96,
"confidence": 0.686
},
{
"text": "can.",
"start": 12.96,
"end": 13.44,
"confidence": 0.19
}
]
},
{
"id": 7,
"seek": 0,
"start": 13.5,
"end": 15.2,
"text": " There's more to be for long.",
"tokens": [
51064,
821,
311,
544,
281,
312,
337,
938,
13,
51114
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.683,
"words": [
{
"text": "There's",
"start": 13.5,
"end": 14.44,
"confidence": 0.519
},
{
"text": "more",
"start": 14.44,
"end": 14.56,
"confidence": 0.811
},
{
"text": "to",
"start": 14.56,
"end": 14.74,
"confidence": 0.418
},
{
"text": "be",
"start": 14.74,
"end": 14.8,
"confidence": 0.95
},
{
"text": "for",
"start": 14.8,
"end": 15.0,
"confidence": 0.987
},
{
"text": "long.",
"start": 15.0,
"end": 15.2,
"confidence": 0.813
}
]
},
{
"id": 8,
"seek": 0,
"start": 15.38,
"end": 16.86,
"text": " What if you can stay as long as you want?",
"tokens": [
51114,
708,
498,
291,
393,
1754,
382,
938,
382,
291,
528,
30,
51214
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.771,
"words": [
{
"text": "What",
"start": 15.38,
"end": 15.54,
"confidence": 0.387
},
{
"text": "if",
"start": 15.54,
"end": 15.62,
"confidence": 0.918
},
{
"text": "you",
"start": 15.62,
"end": 15.7,
"confidence": 0.97
},
{
"text": "can",
"start": 15.7,
"end": 15.98,
"confidence": 0.946
},
{
"text": "stay",
"start": 15.98,
"end": 16.12,
"confidence": 0.959
},
{
"text": "as",
"start": 16.12,
"end": 16.24,
"confidence": 0.307
},
{
"text": "long",
"start": 16.24,
"end": 16.36,
"confidence": 0.993
},
{
"text": "as",
"start": 16.36,
"end": 16.54,
"confidence": 0.992
},
{
"text": "you",
"start": 16.54,
"end": 16.62,
"confidence": 0.96
},
{
"text": "want?",
"start": 16.62,
"end": 16.86,
"confidence": 0.82
}
]
},
{
"id": 9,
"seek": 0,
"start": 16.86,
"end": 17.74,
"text": " My girlfriend.",
"tokens": [
51214,
1222,
10369,
13,
51264
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.141,
"words": [
{
"text": "My",
"start": 16.86,
"end": 17.34,
"confidence": 0.124
},
{
"text": "girlfriend.",
"start": 17.34,
"end": 17.74,
"confidence": 0.159
}
]
},
{
"id": 10,
"seek": 0,
"start": 17.98,
"end": 19.1,
"text": " Really, Missia?",
"tokens": [
51264,
4083,
11,
5275,
654,
30,
51314
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.387,
"words": [
{
"text": "Really,",
"start": 17.98,
"end": 18.34,
"confidence": 0.704
},
{
"text": "Missia?",
"start": 18.7,
"end": 19.1,
"confidence": 0.287
}
]
},
{
"id": 11,
"seek": 0,
"start": 19.72,
"end": 20.22,
"text": " I agree.",
"tokens": [
51314,
286,
3986,
13,
51364
],
"temperature": 0.0,
"avg_logprob": -0.5517418488212253,
"compression_ratio": 1.4345238095238095,
"no_speech_prob": 0.11139103770256042,
"confidence": 0.165,
"words": [
{
"text": "I",
"start": 19.72,
"end": 20.06,
"confidence": 0.129
},
{
"text": "agree.",
"start": 20.06,
"end": 20.22,
"confidence": 0.211
}
]
}
],
"language": "en",
"language_probs": {
"en": 0.8443762063980103,
"zh": 0.0022106480319052935,
"de": 0.003940905444324017,
"es": 0.005020844284445047,
"ru": 0.006651538424193859,
"ko": 0.00397181510925293,
"fr": 0.004679941106587648,
"ja": 0.0019056976307183504,
"pt": 0.0026457991916686296,
"tr": 0.0016556988703086972,
"pl": 0.0017082561971619725,
"ca": 0.0010121084051206708,
"nl": 0.005386579316109419,
"ar": 0.0022808213252574205,
"sv": 0.0031914429273456335,
"it": 0.001183274551294744,
"id": 0.00030389000312425196,
"hi": 0.0005250748945400119,
"fi": 0.0010280467104166746,
"vi": 0.0002745417586993426,
"he": 0.00021381332771852612,
"uk": 0.000865702168084681,
"el": 0.0014161941362544894,
"ms": 0.0030932524241507053,
"cs": 0.0003665613476186991,
"ro": 0.0002008590381592512,
"da": 0.0017216543201357126,
"hu": 0.0004121360252611339,
"ta": 0.0002579081046860665,
"no": 0.0005049595492891967,
"th": 0.0001601387921255082,
"ur": 0.0008132518851198256,
"hr": 7.804541382938623e-05,
"bg": 8.179088763426989e-05,
"lt": 0.00011716003791661933,
"la": 0.004261135123670101,
"mi": 0.003673328086733818,
"ml": 0.00032348925014957786,
"cy": 0.06770503520965576,
"sk": 4.4295567931840196e-05,
"te": 0.00010461213969392702,
"fa": 9.088860679185018e-05,
"lv": 0.0001247162144863978,
"bn": 0.00011716003791661933,
"sr": 4.588053343468346e-05,
"az": 2.771945764834527e-05,
"sl": 0.00023666980268899351,
"kn": 7.146808911784319e-06,
"et": 0.0001042043004417792,
"mk": 2.262396264995914e-05,
"br": 0.0006483809556812048,
"eu": 4.446893581189215e-05,
"is": 0.000752135063521564,
"hy": 1.124350728787249e-05,
"ne": 8.740671910345554e-05,
"mn": 4.210234328638762e-05,
"bs": 0.00017315131844952703,
"kk": 1.1156010259583127e-05,
"sq": 4.922262451145798e-05,
"sw": 0.0002968504268210381,
"gl": 0.0005502737476490438,
"mr": 1.3509384189092088e-05,
"pa": 8.587087904743385e-06,
"si": 3.3962591260205954e-05,
"km": 0.0001786477369023487,
"sn": 0.0002559010754339397,
"yo": 0.0001691402867436409,
"so": 2.7661242256726837e-06,
"af": 0.000508919998537749,
"oc": 7.360383460763842e-05,
"ka": 7.788150469423272e-06,
"be": 9.018131095217541e-05,
"tg": 3.4419832672938355e-07,
"sd": 6.320310785667971e-05,
"gu": 3.510393980832305e-06,
"am": 8.161911864590365e-06,
"yi": 0.00012278267240617424,
"lo": 3.0979122129792813e-06,
"uz": 2.5671094139312345e-09,
"fo": 0.0002499731199350208,
"ht": 0.00021049848874099553,
"ps": 1.2690892617683858e-05,
"tk": 4.414082255266294e-08,
"nn": 0.007898896001279354,
"mt": 4.177469963906333e-05,
"sa": 0.0001042043004417792,
"lb": 5.532596674129309e-07,
"my": 2.5636309146648273e-05,
"bo": 5.491180854733102e-05,
"tl": 0.0005502737476490438,
"mg": 8.147531360691573e-08,
"as": 4.4375542529451195e-06,
"tt": 7.103994335011521e-07,
"haw": 0.004535954911261797,
"ln": 9.806761227082461e-06,
"ha": 1.4645611656760593e-07,
"ba": 4.5977703422295235e-08,
"jw": 0.0009433886734768748,
"su": 8.766671300008966e-08
}
}