VAPI•7mo ago

Error when creating a custom TTS integration

Patching my assistant with the following voice configuration, according to the docs

{
    "voice": {
        "server": {
            "url": "https://WEBHOOK_URL",
            "secret": "SECRET",
            "headers": {
                "Content-Type": "application/json",
                "X-API-Version": "v1"
            },
            "timeoutSeconds": 45
        },
        "provider": "custom-voice",
        "fallbackPlan": {
            "voices": [
                {
                    "model": "gpt-4o-mini-tts",
                    "voiceId": "shimmer",
                    "provider": "openai",
                    "instructions": "Speak in calm and professional tone. "
                }
            ]
        }
    }
}

{
    "voice": {
        "server": {
            "url": "https://WEBHOOK_URL",
            "secret": "SECRET",
            "headers": {
                "Content-Type": "application/json",
                "X-API-Version": "v1"
            },
            "timeoutSeconds": 45
        },
        "provider": "custom-voice",
        "fallbackPlan": {
            "voices": [
                {
                    "model": "gpt-4o-mini-tts",
                    "voiceId": "shimmer",
                    "provider": "openai",
                    "instructions": "Speak in calm and professional tone. "
                }
            ]
        }
    }
}

{
    "voice": {
        "server": {
            "url": "https://WEBHOOK_URL",
            "secret": "SECRET",
            "headers": {
                "Content-Type": "application/json",
                "X-API-Version": "v1"
            },
            "timeoutSeconds": 45
        },
        "provider": "custom-voice",
        "fallbackPlan": {
            "voices": [
                {
                    "model": "gpt-4o-mini-tts",
                    "voiceId": "shimmer",
                    "provider": "openai",
                    "instructions": "Speak in calm and professional tone. "
                }
            ]
        }
    }
}

{
    "voice": {
        "server": {
            "url": "https://WEBHOOK_URL",
            "secret": "SECRET",
            "headers": {
                "Content-Type": "application/json",
                "X-API-Version": "v1"
            },
            "timeoutSeconds": 45
        },
        "provider": "custom-voice",
        "fallbackPlan": {
            "voices": [
                {
                    "model": "gpt-4o-mini-tts",
                    "voiceId": "shimmer",
                    "provider": "openai",
                    "instructions": "Speak in calm and professional tone. "
                }
            ]
        }
    }
}

Cannot access the assistant in the dashboard:

Error: TypeError
Message: Cannot read properties of undefined (reading 'models')

Error: TypeError
Message: Cannot read properties of undefined (reading 'models')

Error: TypeError
Message: Cannot read properties of undefined (reading 'models')

Error: TypeError
Message: Cannot read properties of undefined (reading 'models')

Tried also removing the

secret

secret

field, and add a custom header instead, but without success.

Custom TTS integration | Vapi

Learn to integrate your own text-to-speech system with VAPI

Vapi Support BotAPP•6/19/25, 6:28 AM

It looks like your fallback plan includes a modelmodel field for the openaiopenai provider, but the documentation only shows providerprovider and
voiceId
voiceId
in fallback examples. This may be causing the dashboard error:

TypeError: Cannot read properties of undefined (reading 'models')

TypeError: Cannot read properties of undefined (reading 'models')

Could you try removing the "model": "gpt-4o-mini-tts""model": "gpt-4o-mini-tts" line from your fallback plan and see if the error persists?

Source:

Custom TTS integration

⤵️ Mention @Vapi Support Bot for follow-ups 🐥

static-pinkOP•6/19/25, 6:31 AM

Tried, still the same error.
Btw, in the API reference these fields are allowed in fallbackPlan -> voices

Update Assistant | Vapi

static-pinkOP•6/19/25, 7:01 AM

@Duckie

Vapi Support BotAPP•6/19/25, 7:03 AM

Try using only providerprovider and
voiceId
voiceId
in your fallback plan, as shown below:

"fallbackPlan": {
  "voices": [
    {
      "provider": "openai",
      "voiceId": "shimmer"
    }
  ]
}

"fallbackPlan": {
  "voices": [
    {
      "provider": "openai",
      "voiceId": "shimmer"
    }
  ]
}

Remove any modelmodel or instructionsinstructions fields from the fallback voice. If the error persists, it may be a dashboard bug or an unsupported configuration not covered in the documentation.
See: Custom TTS integration docs

Source:

Custom TTS integration

extreme-purple•6/19/25, 10:31 AM

Add a "models" array under "voice" to prevent the TypeError and ensure Vapi can load the custom voice configuration..

static-pinkOP•6/19/25, 11:45 AM

Hey @Kings_big

,
Thanks for responding.
I get an error, it doesn't expect "models" under "voice".

Payload:

{
    "voice": {
        "models": [ ".." ],
        "server": {
            "url": "https://WEBHOOK_URL",
            "secret": "SECRET",
            "headers": {
                "Content-Type": "application/json",
                "X-API-Version": "v1"
            },
            "timeoutSeconds": 45
        },
        "provider": "custom-voice",
        "fallbackPlan": {
            "voices": [
                {
                    "voiceId": "shimmer",
                    "provider": "openai"
                }
            ]
        }
    }
}

{
    "voice": {
        "models": [ ".." ],
        "server": {
            "url": "https://WEBHOOK_URL",
            "secret": "SECRET",
            "headers": {
                "Content-Type": "application/json",
                "X-API-Version": "v1"
            },
            "timeoutSeconds": 45
        },
        "provider": "custom-voice",
        "fallbackPlan": {
            "voices": [
                {
                    "voiceId": "shimmer",
                    "provider": "openai"
                }
            ]
        }
    }
}

Response:

{
    "message": [
        "voice.property models should not exist"
    ],
    "error": "Bad Request",
    "statusCode": 400
}

{
    "message": [
        "voice.property models should not exist"
    ],
    "error": "Bad Request",
    "statusCode": 400
}

Do you mean something else?

static-pinkOP•6/19/25, 1:22 PM

@Duckie

Vapi Support BotAPP•6/19/25, 1:24 PM

Use only providerprovider and
voiceId
voiceId
in your fallback plan, as shown below:

"fallbackPlan": {
  "voices": [
    {
      "provider": "openai",
      "voiceId": "shimmer"
    }
  ]
}

"fallbackPlan": {
  "voices": [
    {
      "provider": "openai",
      "voiceId": "shimmer"
    }
  ]
}

Remove any modelmodel, instructionsinstructions, or modelsmodels fields from your configuration. If the error persists after this change, it may be a dashboard bug not covered in the documentation.
Source: Custom TTS integration documentation

Source:

Custom TTS integration documentation

extreme-purple•6/19/25, 2:23 PM

Ooh got it, use "voiceId" with "provider": "custom-voice" instead to specify your model..

https://docs.vapi.ai/customization/custom-voices/custom-tts

Custom TTS integration | Vapi

Learn to integrate your own text-to-speech system with VAPI

static-pinkOP•6/19/25, 2:37 PM

Can you write me a request payload for example please?
@Duckie

Vapi Support BotAPP•6/19/25, 2:39 PM

Here’s an example request payload for a custom TTS integration with a fallback plan, following the official documentation:

{
  "voice": {
    "provider": "custom-voice",
    "server": {
      "url": "https://YOUR_TTS_WEBHOOK_URL",
      "secret": "YOUR_SECRET",
      "headers": {
        "Content-Type": "application/json",
        "X-API-Version": "v1"
      },
      "timeoutSeconds": 45
    },
    "fallbackPlan": {
      "voices": [
        {
          "provider": "openai",
          "voiceId": "shimmer"
        }
      ]
    }
  }
}

{
  "voice": {
    "provider": "custom-voice",
    "server": {
      "url": "https://YOUR_TTS_WEBHOOK_URL",
      "secret": "YOUR_SECRET",
      "headers": {
        "Content-Type": "application/json",
        "X-API-Version": "v1"
      },
      "timeoutSeconds": 45
    },
    "fallbackPlan": {
      "voices": [
        {
          "provider": "openai",
          "voiceId": "shimmer"
        }
      ]
    }
  }
}

Only use providerprovider and
voiceId
voiceId
in the fallback voices. Do not include modelsmodels, modelmodel, or instructionsinstructions fields.
See the official docs for reference.

Source:

Custom TTS integration documentation

Eextreme-purple Ooh got it, use "voiceId" with "provider": "custom-voice" instead to specify you...

static-pinkOP•6/19/25, 3:01 PM

Can you write me an example of this payload please?

extreme-purple•6/19/25, 4:02 PM

Set "provider": "custom-voice" with your "voiceId" and provide a "server" object with your TTS endpoint, secret, and headers, plus an optional "fallbackPlan" for backup voices..

embarrassing-maroon•6/28/25, 7:50 PM

Did it work for you? How?

Sstatic-pink Can you write me an example of this payload please?

extreme-purple•6/28/25, 7:57 PM

checking if this is resolved/solved for you?

VapiAPP•6/30/25, 2:56 AM

Hey @lior_cychi, could you please try again with this CURL request.

curl -X POST https://api.vapi.ai/assistant \
     -H "Authorization: Bearer " \
     -H "Content-Type: application/json" \
     -d '{
  "voice": {
    "provider": "custom-voice",
    "server": {
      "timeoutSeconds": 0,
      "url": "",
      "headers": {},
      "backoffPlan": {
        "maxRetries": 0,
        "baseDelaySeconds": 0
      }
    },
    "fallbackPlan": {
      "voices": [
        {
          "provider": "azure"
        }
      ]
    }
  }
}'

curl -X POST https://api.vapi.ai/assistant \
     -H "Authorization: Bearer " \
     -H "Content-Type: application/json" \
     -d '{
  "voice": {
    "provider": "custom-voice",
    "server": {
      "timeoutSeconds": 0,
      "url": "",
      "headers": {},
      "backoffPlan": {
        "maxRetries": 0,
        "baseDelaySeconds": 0
      }
    },
    "fallbackPlan": {
      "voices": [
        {
          "provider": "azure"
        }
      ]
    }
  }
}'

Error when creating a custom TTS integration

Similar Threads

Similar Threads

Similar Threads