Setting the input_device_index has no effect

[TheMancha]

I'm using the file realtimestt_test.py running in a miniconda environment. The script works awesome. The issue I have is that I need it to capture the audio coming from my system, not my mic.

Kolja told me to set the input_device_index in the file using show_devices.py to see the index of the device I want to set. I received 3 different devices. I tried setting the input_device_index to 4, 13 and 22 but none worked. No matter what, it keeps taking the audio coming from my mic.

I know that there's a workaround using stereomix, but that will interfere with other things in how I want to do things. Am I doing something wrong?

This is the device I want the script to take the audio from:

Device Index: 4 Name: FxSound Speakers (FxSound Audio Sample Rate (Default): 44100.0 Hz Max Input Channels: 0 Max Output Channels: 8 Host API: MME

Device Index: 13 Name: FxSound Speakers (FxSound Audio Enhancer) Sample Rate (Default): 44100.0 Hz Max Input Channels: 0 Max Output Channels: 8 Host API: Windows DirectSound

Device Index: 22 Name: FxSound Speakers (FxSound Audio Enhancer) Sample Rate (Default): 48000.0 Hz Max Input Channels: 0 Max Output Channels: 2 Host API: Windows WASAPI

And this is the fragment of the code in realtimestt_test.py that I'm modifiying:

Recorder configuration

recorder_config = {
    'spinner': False,
    'model': 'large-v2', # or large-v2 or deepdml/faster-whisper-large-v3-turbo-ct2 or ...
    'download_root': None, # default download root location. Ex. ~/.cache/huggingface/hub/ in Linux
    'input_device_index': 4,
    'realtime_model_type': 'tiny.en', # or small.en or distil-small.en or ...

[TheMancha]

So, if I use "" in the input device index like:

input_device_index': "4"

I get this error

WARNING:root:Failed to get highest sample rate: 'str' object cannot be interpreted as an integer

Still doesn't get the audio from my system and keeps getting the audio from my mic.

PS. Sorry if this "error" looks stupid, I'm not a programmer myself. I just read and go trial-error until things work. :)

[TheMancha]

I just discovered that stereomix actually works perfect for what I need. So, I'll go that way. I'll leave this open just in case Kolja or other person may propose a solution.

[Nenesh]

Did you set use_microphone = False ? If not, I think it prioritize microphone input. input_device_index take int, so you just do input_device_index = 4