I am trying to build an AI Voice Interviewer and built a persistent Assistant. For some reason even when I try to test in with the "Talk to Assistant" it is not responding. It will recite the first message, and then when I speak back to it I get no response at all. If I chat with the assitant I get a response no problem, but voice isn't working. Can you please help? @Vapi

Hi there! 👋 I understand how frustrating it can be when your voice assistant isn't responding properly. I can definitely help troubleshoot why your persistent Assistant is not responding to voice input after the initial message, even though it works fine with chat. This is actually a common issue with the "Talk to Assistant" feature. Based on the information I've found, here are some potential solutions: ### 1. Adjust Endpointing and Start Speaking Plan One of the most common fixes involves adjusting your transcriber and start speaking plan settings:

json
"start_speaking_plan": {
  "waitSeconds": 0.8,
  "smartEndpointingEnabled": false
},
"transcriber": {
  "model": "nova-2",
  "language": "en",
  "provider": "deepgram",
  "endpointing": 300
}

According to [troubleshooting information](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1333057366748565514), these specific settings have helped many users resolve this exact issue. ### 2. Check Assistant Configuration Make sure: - Your assistant ID is correctly set on the phone number - The "provider" field is present in your assistant overrides and set to "openai" - Your first message is not empty (which can prevent the assistant from speaking initially) ### 3. Consider Transcriber Issues Some users have resolved this issue by changing the transcriber from Deepgram to another provider like Azure. The assistant might not be responding because of connection issues with the transcriber. ### 4. Voice Model Configuration Ensure your voice model language settings are properly configured. If using a custom voice (like from 11labs), verify that it's correctly synced and that you're using your own API key if needed. ### 5. Network and Server Configuration Verify that your server URL is correctly configured so the assistant knows where to send and receive data for processing voice input.

For further clarity, the logs say that the calls are ended due to "Silence timed out" even though when I listen to the recording I can clearly hear myself speaking. My org ID is a995303b-fb4a-41d6-8b90-6d208a26bb33 Thanks.\

Hey! To help track down this issue, could you share: - The call ID This would really help us figure out what went wrong!

@Shubham Bajaj It's literally every one of my calls, whether I do it through the app I built or the "Talk to Assistant" function. Are you able to look at my call logs? Here are a few of the call IDs: Here is one done through the app -- 85f9597b-ff45-489b-92f4-56a514bf253f Here is one done via the "Talk to Assistant" -- f241c6dc-1f38-4e81-9b6a-7ca38dfa4267 I am using Arc as my browser, will try with Chrome and Brave and see if that changes anything. Strangely when I tested the "Talk with Assistant" the first time (in Arc) it worked but not since. Here is the "Talk with Assistant" one that worked initially, though the voice did some super-weird things -- 7892afab-86d8-4020-95cf-c0ae4756ad32 Here's the one from Chrome -- 27b7940e-d7f0-460a-b844-8c7a370daa08 And here's the one from Brave -- c9e48d34-dcfd-4e66-960f-d251b9756dba FYI, I do have video recording turned on too. Not sure if that could be related. Frustrating though. This is consistent across both Assistants I have created (though I only tested v2 via "Talk with Assistant").

Can you try another set of calls by changing your transcriber to Deepgram and setting its endpointing value to 300ms? Also, remove the transcription and pointing plan from your start speaking plan. These changes will allow your assistant transcription models to have ample time to consider user has finished speaking. let me know how it goes for you