While reviewing the logs for recent calls, I noticed that the voice model referenced in the logs doesn’t match the one specified when triggering the call through the VAPI API. Even though the API request explicitly sets a VAPI voice, the logs show a different provider (for example, 11Labs). Could you please check whether this discrepancy is expected behavior or if there might be an issue with how the voice provider is being recorded or routed? Call ID:

019a2bb6-a9eb-744a-96df-13fc7a6e0acb

https://cdn.discordapp.com/attachments/1433473815714725899/1433473815912124597/Screenshot_2025-10-30_at_10.05.44_AM.png?ex=6904d206&is=69038086&hm=1a07409e5d0e11378dbb1e09ef33710fa6e699033517b292ce5c325b9d27c486& https://cdn.discordapp.com/attachments/1433473815714725899/1433473816394465481/Screenshot_2025-10-30_at_10.07.42_AM.png?ex=6904d206&is=69038086&hm=d2e3998f6a54a7068db1407d735e75b62258925b339f01d121f9ee4f9995965f&

This behavior is likely due to Vapi's Voice Fallback Plan, which automatically switches to a backup provider (like ElevenLabs) if the primary voice (e.g., "Paige") is unavailable. To check or modify your fallback configuration, review your assistant setup for a section like this:

json
"voice": {
  "provider": "vapi",
  "voiceId": "Paige",
  "fallbackPlan": {
    "voices": [
      { "provider": "elevenlabs", "voiceId": "dOPyu4KyOAHZJaFmb3" }
    ]
  }
}

Action: Review your assistant's fallback plan to confirm if this is expected, and adjust as needed. For more details, see the [Voice Fallback Plan documentation](https://docs.vapi.ai/voice-fallback-plan.mdx). Source: - [Voice Fallback Plan documentation](https://docs.vapi.ai/voice-fallback-plan.mdx)

Hey @Vapi Support Bot , there 's no fallback plan in place. Here is the voice config from the assistant setup:

"voice": {
  "voiceId": "Elliot",
  "provider": "vapi",
  "inputPunctuationBoundaries": [
    "||"
  ]
},

and more importantly, here is the voice configuration from the assistant overrides:

"voice": {
  "voiceId": "Paige",
  "provider": "vapi"
},

While reviewing other call details, I noticed that there are costs associated with both Vapi and 11Labs. This confirms that

11Labs

is being used in addition to

Vapi

, even though no fallback is configured:

costs: [
  ...
  {
    "cost": 0.02815,
    "type": "voice",
    "voice": {
      "model": "eleven_turbo_v2_5",
      "voiceId": "dOPyuP4KyOAHZjVaFmb3",
      "provider": "vapi"
    },
    "characters": 563
  },
  {
    "cost": 0.11731,
    "type": "vapi",
    "minutes": 2.3462,
    "subType": "normal"
  },
  ...
]

Hi rmaceissoft, If you encounter voice model errors, ensure that parameters like

provider

,

voiceId

, and

model

are correctly specified for your chosen voice, such as those from Cartesia, ElevenLabs, or any other voice provider. Review the fallback plans to provide alternative voice options if primary ones fail. This can enhance the reliability of voice sessions. If issues persist, consider adjusting properties such as

speed

or

language

to ensure compatibility with your application needs. For more detailed setup or parameter configuration, refer to the relevant [API documentation](https://docs.vapi.ai/api-reference/sessions/list).