Speech Recognition with Virtual Cable Microphone Distorted.

[timovr-dev]

I use speech_recognition with a Virtual Cable (https://vb-audio.com/Cable/). The issue is speech_recogniton captures a very distorted sound regardless if I speak or not. Recorded Sound [Headphone warning]: https://youtu.be/q78puE0Tdjc

However, this only happens with the speech_recognition package. If I use the Virtual Cable and use it in OBS as a microphone I hear everything perfectly fine. I also checked the sample rate and 44100 Hz seems to be correct.

Code:

def record_audio(audio_queue, energy, pause, dynamic_energy, save_file, temp_dir):
    r = sr.Recognizer()
    r.pause_threshold = 2
    r.non_speaking_duration = 1
    microphone_name = "CABLE-B Output (VB-Audio Cable B)"
    microphone_index = find_index(microphone_name)

    with sr.Microphone(device_index=microphone_index, sample_rate=44100) as source:
        r.adjust_for_ambient_noise(source, duration=10)
        r.dynamic_energy_threshold = True  
        save_file = True
        while True:
            #get and save audio to wav file
            audio = r.listen(source) 
            if save_file:
                data = io.BytesIO(audio.get_wav_data())
                audio_clip = AudioSegment.from_file(data)
                filename = f"./temp.wav"
                audio_clip.export(filename, format="wav")

System Information: Windows 10, Python 3.9, speech_regonition==3.10.0