WebRTC send-receive audio creates High CPU Utilization

[BajaMexico]

If you run the code from https://fastrtc.org/userguide/gradio/ for the WebRTC Component it creates a 100% utilization of the CPU. This happens as soon as you press "Record" on the UI, there doesn't need to be any audio to actually call the ReplyOnPause method. The fans start to really go fast to keep the CPU cool. And while the example is fairly trivial if you do anything in your ReplyOnPause method (I am using Meta Seamless) the actual audio response starts to talk really slow. I assume this is due to the high CPU utilization-> high fans-> thermal throttling of CPU, Am I doing something wrong, is there a way to keep the CPU utilization to a lower value, is this just the nature of monitoring for audio input?

This is with:
fastrtc 0.0.33
fastrtc-moonshine-onnx 20241016

on a Windows 11 WSL2 stock linux system

Here is the code from the webpage example above with additions of the line
import numpy as np

and line change of
demo.launch(server_name="127.0.0.1", server_port = 49000)

import gradio as gr
import numpy as np
from fastrtc import WebRTC, ReplyOnPause

def response(audio: tuple[int, np.ndarray]):
"""This function must yield audio frames"""
...
yield audio

with gr.Blocks() as demo:
gr.HTML(
"""

Chat (Powered by WebRTC ⚡️)

"""
)
with gr.Column():
with gr.Group():
audio = WebRTC(
mode="send-receive",
modality="audio",
)
audio.stream(fn=ReplyOnPause(response),
inputs=[audio], outputs=[audio],
time_limit=60)
demo.launch(server_name="127.0.0.1", server_port = 49000)

[BajaMexico]

What I don't understand is that if you use the same code but set the AlgoOptions and the SileroVadOptions, specifically the audio_chunk_duration in the AlgoOptions, it doesn't affect anything. I changed it to 5.0 seconds, which I assume means that the Vad runs every 5 seconds. This still results in an overclocked CPU. Of course maybe my assumption is wrong, but it seems like the audio_chunk_duration has no affect on the periodicity of the Vad algorithm.

Here is my code that I modified.

import gradio as gr
import numpy as np
from fastrtc import Stream, ReplyOnPause, WebRTC, AlgoOptions, SileroVadOptions

def response(audio: tuple[int, np.ndarray]):
"""This function must yield audio frames"""
...
yield audio

with gr.Blocks() as demo:
gr.HTML(
"""

Chat (Powered by WebRTC ⚡️)

"""
)
with gr.Column():
with gr.Group():
audio = WebRTC(
mode="send-receive",
modality="audio",
)
#audio.stream(fn=ReplyOnPause(response),
#inputs=[audio], outputs=[audio],
#time_limit=60)
audio.stream (fn=ReplyOnPause (response, algo_options=AlgoOptions( audio_chunk_duration=5.0,
started_talking_threshold=0.3, speech_threshold=0.2), model_options=SileroVadOptions( threshold=0.5,
min_speech_duration_ms=250, min_silence_duration_ms=250), output_sample_rate=16000),
inputs=[audio], outputs =[audio], time_limit=3600)

demo.launch(server_name="127.0.0.1", server_port = 49000)