I'm not using the Realtime audio API here - I'm using the Chat Completion API, which added support for audio files a few days ago: https://simonwillison.net/2024/Oct/18/openai-audio/