### Answer Issue 1: Repeating Phrases For the repeating phrase issue, this often relates to your prompt structure. Based on the video guidance, you should explicitly instruct the agent in your prompt about when and how to use filler phrases [(1) ]() . Consider adjusting your agent's prompt to be more specific about when to say "one moment" and ensure it's not being triggered multiple times during function calls . *Issue 2: Silent Periods During Function Calls (Latency) * The 2-3 second silence during function calls is a latency issue. You can diagnose this by checking the latency breakdown in your call dashboard or using the Get Call API [(2) ]() . The latency object will show you which component is causing the delay : -

e2e

- End-to-end latency from when the user stops talking to when the agent starts talking -

llm

- LLM latency from the start of the LLM call to the first speakable chunk received -

tts

- Text-to-speech latency How to reduce latency: If LLM inference is the bottleneck, switch to a faster model [(3) ]() . If TTS is the bottleneck, choose a lower-latency voice provider . If tool calls are slow, optimize tool endpoints or reduce response size . You can also adjust agent responsiveness settings. The

responsiveness

parameter controls how responsive the agent is, with values ranging from 0 to 1 [(4) ]() . Higher values mean faster exchanges .

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.

​ To learn more, see the sources I considered: [[1] Master The Future of Voice AI - Conversation Design & Agent Persona]() [[2] Check actual latency](https://docs.retellai.com/reliability/check-actual-latency) [[3] Address Metric Issues](https://docs.retellai.com/ai-qa/address-metric-issues) [[4] Create Voice Agent](https://docs.retellai.com/api-references/create-agent) ​

Thanks for the follow-up, @flowrpot. I'll get back to you soon.

### Answer Based on your issues, here are the recommended changes: For the repeating "one moment" phrase: Configure your custom function's Speak During Execution setting [(1) ]() . When this is enabled, the agent speaks at the moment the function is called . You should also update your prompt to explicitly state when to use filler phrases and ensure it only speaks once during function execution . For the 2-3 second silence during function calls: 1. Check your latency breakdown using the Get Call API to identify which component is slow (LLM, TTS, or the custom function itself) [(2) ]() . The response includes fields like

e2e

,

llm

,

tts

, and statistical metrics (p50, p90, min, max) . 2. **Adjust Responsiveness**: Set the responsiveness lower if you want the agent to wait longer before responding [(3) ]() . You can also enable "Dynamically adjust based on user input" to let the agent automatically tune its response timing during the call . 3. **Configure function speech behavior**: In your custom function settings, ensure "Speak after execution" is enabled so the agent continues talking after the function completes without waiting for user input [(1) ]() . This prevents awkward silences . 4. **Optimize your custom function endpoint**: The request will timeout in your specified timeout period, or 2 minutes if not specified . Make sure your endpoint responds quickly to avoid delays. 5. **Consider using backchanneling**: Configure how often and what words the agent uses to acknowledge users in the Speech Settings section [(3) ](

​ To learn more, see the sources I considered: [[1] Integrate any system with custom function](https://docs.retellai.com/build/single-multi-prompt/custom-function) [[2] Check actual latency](https://docs.retellai.com/reliability/check-actual-latency) [[3] Step 2: Configure the basic settings](https://docs.retellai.com/build/single-multi-prompt/configure-basic-settings) ​

Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.