ReSpeaker XVF3800 microphone does not work with PyAudio, Can you create an arecord implementation?

[BryanSwie]

The XVF3800 is a 4-mic array and only exports a 6-channel stream at the ALSA level (4 raw mics + 2 DSP-processed channels). PyAudio/SpeechRecognition asks for 1 channel, ALSA rejects it.

What works is: arecord plughw:2,0 -c 6 captures a fixed-duration WAV with all 6 channels

Python reads the raw PCM, reshapes it with numpy into (frames × 6), and slices out channel 0

That mono buffer is wrapped in sr.AudioData in my code and sent to Google STT directly for use on a local LLM running on a Hailo H10 Device attached to a Raspberry Pi 5.

[frymanofer]

Hi Bryan - thank you for the report.
We create:
Python_WakeWordDetection/example/example_external_audio.py
example so you can control the audio you feed to the model.
Can you check if you can use the API's in this eample?
Tx.