how is the best option to capture user audio event...
# ask-ais
salmon-fountain-5704
03/26/2025, 4:00 PMhow is the best option to capture user audio events?
t
tall-airline-16756
03/26/2025, 4:00 PMThanks for reaching out, Pablo. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
tall-airline-16756
03/26/2025, 4:00 PMTo capture user audio events, you have two main options:
• Server-side Voice Activity Detection (VAD): This is the default option where you can configure parameters like threshold, prefix padding, and silence duration. It is suitable for general use cases.
• Client-side VAD: This option provides more control over audio input and is particularly useful for push-to-talk interfaces. Note that this is currently available only for Python.
Additionally, you can listen for specific system events such as
user_started_speakinguser_stopped_speakinguser_speech_committeds
salmon-fountain-5704
03/26/2025, 4:13 PMcoul you helpme creating the code ?
t
tall-airline-16756
03/26/2025, 4:14 PMHere is an example code snippet to help you capture user audio events using Voice Activity Detection (VAD):
Copy code
from livekit import agents, rtc
from livekit.plugins import openai, silero
async def process_track(ctx: agents.JobContext, track: rtc.Track):
# Initialize Whisper STT
whisper_stt = openai.STT()
# Configure VAD with specific parameters
vad = silero.VAD.load(
min_speech_duration=0.1,
min_silence_duration=0.5,
)
vad_stream = vad.stream()
# StreamAdapter buffers audio until VAD emits END_SPEAKING event
stt = agents.stt.StreamAdapter(whisper_stt, vad_stream)
stt_stream = stt.stream()
# ...user_started_speakinguser_stopped_speaking