[Transcriptions] Changes between providers #178
Description
I had a quick question before syncing some time into getting the Elevenlabs transcriptions to behave.
Currently there seems to be a change between how the Elevenlabs gateway and the Prism gateway (the one used for OpenAI and Mistral) handle transcribing content.
When transcribing content the OpenAI model (and most likely the mistral model) return full sentences, whereas elevenlabs returns a word by word basis.
Would this be something that's supposed to be "fixed" within the package (making all the gateways/providers unified by changing the Elevenlabs gateway to follow the response format of prism) or would it be up to the application to make this change (which would make changing models a bit harder for quick debugging/testing)