Here is the output I return:

{"assistantId":"ASSISTANT_ID","phoneNumberId":"PHONE_NUMBER_ID","assistantOverrides":{"variableValues":{"firstName":"George"},"voice":{"provider":"vapi","voiceId":"hana"}}}

It doesn't change any of the settings.

To override the assistant voice for inbound calls, try using a supported provider like

"azure"

in your

assistantOverrides

. For example:

json
"assistantOverrides": {
  "voice": {
    "cachingEnabled": true,
    "provider": "azure",
    "voiceId": "andrew"
  }
}

Make sure the provider and voiceId are valid and supported for your account, as using

"vapi"

may not work for this override according to the documentation I have found. See the API reference for more details: [Vapi API Reference](https://docs.vapi.ai/api-reference/calls/create-phone-call). Source: - [Vapi API Reference: Create Phone Call](https://docs.vapi.ai/api-reference/calls/create-phone-call)

Is there any chance I can override the voice with Vapi voice?

I can help you get the Vapi voice overrides working correctly. It looks like your request reaches the endpoint, but assistantOverrides aren’t being applied. Often, this happens if the request is sent before a conversation starts or in the wrong context. Can you confirm whether you’re sending this during an active conversation or as a pre-call setup? Once we clarify, I can assist you on the correct request structure so the voice changes actually take effect. @Aram Vardanyan

Actually, I'm new to Vapi and didn't know how to handle pre-call setups so it can maintain the assistant config and override the voice only. Now, I just removed the attached assistant from the phone number and it worked. Seems the problem was that overriding was disabled if there is an assistant attached.

I’m glad you got the override issue sorted — that makes perfect sense when an assistant is attached to the phone number. High latency usually comes from a mix of model size, TTS rendering speed, and how your pre-call/webhook logic is structured. I can help you review your setup and optimize things like model choice, streaming, server response times, and Vapi fast-start patterns. With the right configuration, we can cut your latency down significantly and smooth out the call experience. If you’d like, we can go over the details privately so I can assist you @Aram Vardanyan ?

Hi there, Thanks for reaching out! To help us review why the voice override isn’t being applied on inbound calls, could you please share an example Call ID where this happened? Once we have that, we’ll take a closer look at the configuration and identify what needs to be adjusted. Feel free to forward just one recent call ID and we’ll handle the rest. Looking forward to your reply!