I have been trying out Gladia to seem if it potentially provides better transcribing performance than Deepgram. A call came in where the transcriber only transcribed the assistant messages, not the caller's. But this doesn't happen every time, some calls came in with the same exact setup and went perfectly fine. This is the call information for the failed call. Call ID: 84934bfd-17ec-4372-ac6c-b6d6a5e9c0f2 Time: 10:11AM EST Is this a Gladia issue? Is it just with the Solaria model? Was there an error in the code in the way it was configured? Something else?

Hi Lightheaded, The issue you're encountering with the Gladia transcriber may be related to a specific setup or model issue, potentially with how Gladia handles certain calls. Since you've noted inconsistencies in performance, it might be beneficial to check your current usage to ensure you haven't hit any limits similar to past issues with the Gladia service, as detailed in earlier cases such as hitting the free limit. Here are a few steps you can explore: 1. Verify your usage limits and ensure your API key is active and within the allowable limits. 2. Review if the Solaria model is involved, which might have unique issues. 3. Consider setting up a fallback transcriber to manage inconsistent results. See if any commonalities exist among calls where the transcriber worked compared to where it didn't, like specific call conditions or usage patterns. If this persists, assessing the system's logs or contacting Gladia's support might provide further insights. For detailed information on the error codes and potential causes, refer to the [call ended reasons document](https://docs.vapi.ai/calls/call-ended-reason).

We aren't using any Gladia products separately. We are just using choosing Gladia from the dropdown menu of transcriber options in Vapi

We are seeing this setting in the transcriber:

"transcriber": {
          "model": "solaria-1",
          "prosody": false,
          "language": "en",
          "provider": "gladia",
          "languages": [
            "en",
            "es"
          ],
          "audioEnhancer": false,
          "confidenceThreshold": 0.4
        },

The languages array may be conflicting with each other and it would be best to remove

es

to eliminate any confusion for the assistant.