Invalid voice.model value when using Cartesia snapshot (sonic-3-2025-10-27) Vapi AI #support

Invalid voice.model value when using Cartesia snap...

rmaceissoft

11/04/2025, 7:16 AM

When attempting to configure my assistant to use the Cartesia provider with a specific snapshot for consistency (sonic-3-2025-10-27), the API returns a 400 Bad Request error indicating that the voice.model value is invalid. Request: Please confirm whether Cartesia snapshots (e.g. sonic-3-2025-10-27) are supported, or if only base model names are currently allowed. Reproduction:

Copy code

curl -X PATCH https://api.vapi.ai/assistant/<ASSISTANT_ID> \
  -H "Authorization: Bearer <API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "voice": {
      "provider": "cartesia",
      "voiceId": "a0e99841-438c-4a64-b679-ae501e7d6091",
      "model": "sonic-3-2025-10-27"
    }
  }'

Response:

Copy code

{
  "message": [
    "voice.model must be one of the following values: sonic-3, sonic-2, sonic-english, sonic-multilingual, sonic-preview, sonic"
  ],
  "error": "Bad Request",
  "statusCode": 400
}

Expected Behavior: The API should accept valid Cartesia model snapshots (e.g., sonic-3-2025-10-27) if they are supported, or documentation should clarify the allowed values for voice.model.

Vapi Support Bot

11/04/2025, 7:17 AM

Cartesia snapshot model names (e.g.,
sonic-3-2025-10-27
) are not currently supported. You must use one of the base model names for

voice.model

, such as

sonic-3

. For example:

Copy code

json
"voice": {
  "provider": "cartesia",
  "voiceId": "a0e99841-438c-4a64-b679-ae501e7d6091",
  "model": "sonic-3"
}

For the latest supported values, refer to the [Cartesia provider documentation](https://docs.vapi.ai/providers/voice/cartesia). Source: - [Cartesia Provider Documentation](https://docs.vapi.ai/providers/voice/cartesia)

Jeebs from eva.be

11/04/2025, 5:27 PM

@rmaceissoft on my end it weirdly cuts the end of sentences with sonic-3. I mean that the full audio is not said by the agent, it cuts the end of the last sentence. Do you have the same issue ?

rmaceissoft

11/04/2025, 6:15 PM

@Jeebs from eva.be so far I haven’t experienced that issue. However, I’m still using it in a test environment. This ticket was mainly to unblock us before starting to use it in production.

Chiranjeet Mishra

11/06/2025, 2:12 AM

Here is the link to the documentation, which includes all accepted enums for the model: Basically we don't accept specific snapshots for Cartesia

rmaceissoft

11/06/2025, 4:22 AM

Thanks for pointing that out! Is there any plan to support Cartesia snapshot names in the near future? Relying on the base model name in production isn’t ideal, since it can introduce inconsistencies when the underlying model updates.

Chiranjeet Mishra

11/08/2025, 1:03 AM

At the moment, support for Cartesia snapshot model names isn’t on our roadmap. That said, feel free to submit it as a feature request so our team can review it for future updates.

rmaceissoft

11/10/2025, 10:12 PM

thx! I went ahead and submitted a feature request[0]. In the meantime I'll be using a custom TTS integration as suggested. [0] - https://vapi.canny.io/feature-requests/p/support-cartesia-snapshot-names-in-voicemodel

Chiranjeet Mishra

11/12/2025, 9:30 PM

Thanks for submitting it! Our team looks at the roadmap pretty often so this should help gain some traction with the community

2 Views

Previous Next