Have any of you tried getting the metering levels (soundwaves) out of assistant's playback? They could be used dynamically manipulate values that looks like the AI's voice matches the brightness of its core (example: https://vapi.ai?demo=true&shareKey=02fca22c-5651-4454-b7e9-e0d3ca5d0110&assistantId=5ee7881e-5860-4971-bb59-332df5d8a9da)