modelOutputInMessagesEnabled set to true but still...
# support
a
Hi, i set "modelOutputInMessagesEnabled": true but I still see the transcription of assistant’s speech inside the conversation history instead of model’s output. see call ID fb941981-c42f-45db-95f1-6898f164524f, all conversation history messages don't include diacitic signs but many messages returned with sign from the LLM (Azure OpenAI). few messages includes other words as well. for example the mesgage on exact local time 2025-08-11 23:15:54.327 the LLM returned "לא בדיוק, בֵּלָה. נסי לחזור..." but the message history show "לא בדיוק בלה נסיא לחזור...". I tried to set the modelOutputInMessagesEnabled on both assistant level and assistantOverrides (see call id d7483629-f52f-4ff4-855a-9638a26db285) still not working. I tried set the modelOutputInMessagesEnabled only on assistantOverrides still not working. Please help.
c
Hi Amir76, The setting
"modelOutputInMessagesEnabled": true
should reflect the model's output instead of the transcription in your conversation history. As this feature is now generally available [as of June 20, 2025](https://docs.vapi.ai/changelog/2025/6/20), ensure that your assistant configuration is up-to-date with this setting. If the issue persists, double-check that the setting is correctly placed within the assistant configuration. It might also be worthwhile to verify that there are no custom configurations or overrides at other nodes or levels that could impact this setting. If these steps don't resolve the issue, consider revisiting any potential conflicts with other configuration settings or consult the changelog for any recent updates that may influence this behavior.
a
Is it only works for 11labs TTS or this settings should work with OpenAI / Azure TTS as well?
c
Sorry about that Amir,
modelOutputInMessagesEnabled
 only works for 11labs voices
a
OK, not good for me as no Hebrew support for 11labs TTS. I suggest you will update Vapi docs to reflect it is only for 11labs.
c
While it is possible to configure for other voice providers, it would require extensive reworking of some infrastructure code. If you are looking for this functionality with another provider, is there one that you prefer?
a
Hi Kyle, sorry for my late reply. I would prefer OpenAI voice provider. this is the provider we are using. Tnx
c
No problem—thanks for confirming OpenAI as your voice provider. If you run into any other issues, feel free to reach out anytime.
3 Views