on_realtime_transcription_stabilized continuously repeating callbacks

[jacobtang]

Hi,when I use realtime transcription call back for a period of time,text_detected will always return the same words，even though there is no real-time voice data input.
recorder_config = {
'spinner': False,
'use_microphone': False,
'model': 'distil-small.en',
'language': 'en',
#'debug_mode': True,
#'level': logging.INFO,
#'silero_sensitivity': 0.4,
#'webrtc_sensitivity': 0,
#'post_speech_silence_duration': 0.1,
#'min_length_of_recording': 1,
#'min_gap_between_recordings': 0.1,
'enable_realtime_transcription': True,
#'realtime_processing_pause': 0.5,
'realtime_model_type': 'distil-small.en',
'beam_size_realtime': 1,

    'on_realtime_transcription_stabilized': text_detected,
}

[KoljaB]

That’s the expected behavior. Since you're using use_microphone=False, it's likely that you're feeding audio chunks from an external source. The RealtimeSTT model processes full sentences, and it reports the full text of a sentence from the beginning when using the real-time model. This is necessary because speech to text models rely on context to accurately transcribe.

If no voice data is being inputted, RealtimeSTT doesn't process anything new and thus returns the same text as before. This is the intended behavior and works as expected. Do you think it should function differently and if so, how would you think it should work instead?