Not an expert, but my suspicion is that the camera following lips can add an extra streaming data point making transcription accuracy much higher even at low vo

Not an expert, but my suspicion is that the camera following lips can add an extra streaming data point making transcription accuracy much higher even at low volumes. Again a hunch and I guess the computational power and battery needs might still be insurmountable