Whisper API verbose_json results

I submitted an audio file to the Whisper API of nonsense words and asked for the results as verbose_json.

Below was the data returned. Are there any API docs available that describe all of the data types returned? I am trying to determine how I can use this data.

Thank you.

{
    "task": "transcribe",
    "language": "english",
    "duration": 9.28,
    "segments": [
        {
            "id": 0,
            "seek": 0,
            "start": 0.0,
            "end": 10.6,
            "text": " Fubar, Malfo, Beatrix, Sing Nog, Klet, Fark.",
            "tokens": [
                50364,
                479,
                836,
                289,
                11,
                376,
                1678,
                78,
                11,
                16031,
                6579,
                11,
                7474,
                426,
                664,
                11,
                591,
                2631,
                11,
                479,
                809,
                13,
                50894
            ],
            "temperature": 0.0,
            "avg_logprob": -0.522584597269694,
            "compression_ratio": 0.8461538461538461,
            "no_speech_prob": 0.34786054491996765,
            "transient": false
        }
    ],
    "text": "Fubar, Malfo, Beatrix, Sing Nog, Klet, Fark."
}
1 Like