We're currently using deepgram for all of our STT work, but have had struggles with it's multilingual capabilities. Gladia seems to do better and we'd potentially want to use them. Is that on the roadmap for inference?

Ooc, which issues did you face?

We don't have an problems detecting the language, but with Spanish we've not had good luck getting correct transcripts. When we take the same audio clip and pass it through the mono lingual spanish, it performs just fine. When we run it through Gladia multilingual it also transcribes correctly.

Thank you for letting me know about what providers you would like in our inference service. We're still working on plans for which providers to include in the future. You can still use gladia on LiveKit by using the gladia plugin. https://docs.livekit.io/agents/models/stt/plugins/gladia/

+1 for Gladia (and Soniox) via Inference. They seem to work best for our multilingual use cases.