Suppress arbitrary speech in silence

[shivamidow]

There are similar posts (e.g., #735 and #567) to this question, but I could not find a solution for this issue.

I want to disable or at least suppress non-speech output. How can I do that?

I already tried whisperParams.suppress_blank = true and whisperParams.suppress_non_speech_tokens = true. While the blank message and onomatopoeic words were gone with the options, Whisper.cpp started outputting arbitrary speech even if the input was empty (i.e., silent noise). The following is part of the random talks that I got. Where do they come from, and why? And how can I completely disable them?

init: found 3 capture devices:
init:    - Capture device #0: 'USB Advanced Audio Device Mono'
init:    - Capture device #1: 'Built-in Audio Analog Stereo'                         
init:    - Capture device #2: 'Studio Display Mono'
init: attempt to open default capture device ...
init: obtained spec for input device (SDL Id = 2):
init:     - sample rate:       16000
init:     - format:            33056 (required: 33056)
init:     - channels:          1 (required: 1)  
init:     - samples per frame: 1024
whisper_init_from_file_no_state: loading model from 'models/ggml-base.en.bin'
whisper_model_load: loading model
whisper_model_load: n_vocab       = 51864
whisper_model_load: n_audio_ctx   = 1500
whisper_model_load: n_audio_state = 512
whisper_model_load: n_audio_head  = 8
whisper_model_load: n_audio_layer = 6
whisper_model_load: n_text_ctx    = 448
whisper_model_load: n_text_state  = 512
whisper_model_load: n_text_head   = 8                                                
whisper_model_load: n_text_layer  = 6
whisper_model_load: n_mels        = 80
whisper_model_load: ftype         = 1
whisper_model_load: qntvr         = 0
whisper_model_load: type          = 2
whisper_model_load: mem required  =  310.00 MB (+    6.00 MB per decoder)
whisper_model_load: adding 1607 extra tokens
whisper_model_load: model ctx     =  140.66 MB
whisper_model_load: model size    =  140.54 MB
whisper_init_state: kv self size  =    5.25 MB
whisper_init_state: kv cross size =   17.58 MB

main: processing 48000 samples (step = 3.0 sec / len = 10.0 sec / keep = 0.2 sec), 4 threads, lang = en, task = transcribe, timestamps = 0 ...
main: n_new_line = 2, no_context = 1

 Okay? I'm going to do a little bit of this.
 Take it out if you think about it, and then do it first.
 I'm going to go home!                                                               
 I'm not going to do this. I'm not going to do this. I'm not going to do this. I'm not going to do this. I'm not going to do this. I'm not going to do this.
 You                              
 Thanks for watching!                                                                
 No. No. The other way around where we are, is walk.
 You
 You
 Ahhh.                                                                                                                                                                     ♪♪                                                                                  
 you
 We will see you next time.                                                          
 You
 and we'll see you in the next one.
 ...

[sdip15fa]

I have thought of a better solution that works perfectly. Remove the silence using ffmpeg. See https://stackoverflow.com/questions/25697596/using-ffmpeg-with-silencedetect-to-remove-audio-silence.