For assistants that use models like openai-4o and gemini-2.0 that have voice to text transcribers built inside the model, what happens to selected transcriber option, is it disregarded/ignored and it defaults to transcriber baked in llm? @Kings_big💫 @Shubham Bajaj