To manually trigger a user speech event on a Python server using the LiveKit realtime API with Voice Activity Detection (VAD) turned off, you should disable server-side VAD by setting

turn_detection=None

in your model or agent configuration. Once VAD is disabled, you can explicitly trigger speech responses using the

generate_reply

method. The

generate_reply

method will: • Cancel any pending generation • End the current user activity if active • Add a placeholder user turn, which is required for Gemini-based models Example:

model = openai.realtime.RealtimeModel(
    voice="alloy",
    temperature=0.8,
    instructions="You are a helpful assistant",
    turn_detection=None,  # Disables server-side VAD
)
agent = multimodal.MultimodalAgent(model=model)
agent.start(ctx.room)

# When you want to trigger speech manually:
agent.generate_reply(...)

For more advanced manual control (such as push-to-talk), you can use the

AgentSession

class with

turn_detection="manual"

. This provides access to methods like

start_user_activity()

,

session.clear_user_turn()

, and

session.commit_user_turn()

to precisely manage when speech events are triggered, ideal for applications requiring programmatic activation. Sources: Turn detection | LiveKit Docs | livekit-agents/livekit/agents/voice/agent_session.py | Turn detection and interruptions | LiveKit Docs