https://livekit.io logo
Join Slack
Powered by
I want my agent to start speaking in an outbound c...
# ask-ai
j
I want my agent to start speaking in an outbound call after 1.5 seconds. How can I achieve it?
✅ 1
t
Thanks for reaching out, Juan. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Thanks — I understand you want the agent to start speaking 1.5 seconds after an outbound call begins. You can do this by adding a short delay in your agent logic and then calling the speech method. Recommended approach: 1. Wait for the outbound participant to join the room (e.g., use your job/session hook like 
ctx.wait_for_participant()
). 2. Pause for 1.5 seconds, then invoke the agent speech: ◦ Python: 
await asyncio.sleep(1.5)
then 
await session.say("Your greeting...")
◦ Node.js: 
await new Promise(r => setTimeout(r, 1500))
then 
await session.say("Your greeting...")
1. To reduce TTS latency, pre-synthesize and cache the audio at startup and pass that audio to 
session.say()
instead of raw text. Notes: • 
session.say()
requires a TTS plugin if you pass text; pre-synthesized audio avoids that dependency and plays faster. • Place the delay after you confirm the SIP/participant has joined so the timing aligns with the call start. If you’d like, tell me which runtime you’re using (Python or Node), and I’ll provide a short code snippet tailored to your environment. Sources: Agent speech and audio | LiveKit Documentation | External data and RAG | LiveKit Documentation | Job lifecycle | LiveKit Documentation | Audio customization | LiveKit Documentation, with interactive elements
j
I use Python and it didn't work
t
Thanks — sorry that didn't work. To help diagnose, could you paste the Python snippet you used and any error output or logs? Also confirm: • Are you waiting for the outbound participant to join (e.g., using 
ctx.wait_for_participant()
or similar) before the delay? • Where did you place 
await asyncio.sleep(1.5)
(inside the same async function that calls 
session.say
)? • Are you passing text to 
session.say()
(which needs a TTS plugin) or pre-synthesized audio? If you want, I can share a minimal working example for Python once you confirm those details. , with interactive elements
j
I placed it after the ctx.wait_for_participant. it is juste before await.session.start() and it didn't work
t
Thanks — appreciate the details. A couple quick things to check and a corrected minimal flow. Common issues I see: • Typo: it should be 
await session.start()
(no dot between 
await
and 
session
). • Order matters: start the session first, then wait 1.5s, then call 
session.say()
. • If you pass text to 
session.say()
you need a TTS plugin; otherwise pre-synthesize audio and pass it. Minimal correct sequence (outline): 1. Wait for participant to join: 
await ctx.wait_for_participant(...)
2. Start the session: 
await session.start()
3. Delay: 
await asyncio.sleep(1.5)
4. Speak: 
await session.say("Hello...")
(or pass pre-synthesized audio) If you share the exact snippet you used (including where 
await.session.start()
appears), I’ll point out the specific fix and provide a runnable Python example. , with interactive elements
2 Views
Open in Slack
PreviousNext
Powered by