Add empty audio catch

[chainyo]

Here is an example of an empty audio returned to a request:

{
    "utterances": [
        {
            "text": "Empty audio",
            "start": 0.0,
            "end": 27.52,
            "speaker": null,
            "words": null
        }
    ],
    "audio_duration": 27.52,
    "alignment": false,
    "num_speakers": -1,
    "diarization": false,
    "source_lang": "en",
    "timestamps": "s",
    "vocab": [],
    "word_timestamps": false,
    "internal_vad": false,
    "repetition_penalty": 1.2,
    "compression_ratio_threshold": 2.4,
    "log_prob_threshold": -1.0,
    "no_speech_threshold": 0.6,
    "condition_on_previous_text": true,
    "process_times": {
        "total": 0.0,
        "transcription": 0.0,
        "diarization": null,
        "alignment": null,
        "post_processing": 0.0
    },
    "dual_channel": false
}

[aleksandr-smechov]

Let's do <EMPTY AUDIO>

Also why num_speakers -1 instead of 0?

[chainyo]

Let's do <EMPTY AUDIO>

Changes are made.

Also why num_speakers -1 instead of 0?

@aleksandr-smechov This is the API request parameter, not the number of detected speakers. There is no features to extract the number of detected speaker yet