Support voice custom model with different endpoint like whisper-1, tts-1, tts-1-hd. and possibilities to set up it instead of Web API and hard coded Openai Whisper