What is the limit for the length of the audio files with the new API (https://platform.openai.com/docs/api-reference/audio/createTranscription) with gpt-4o-transcribe or gpt-4o-mini-transcribe? Or what is the ideal length of the audio files, e.g. for a whole meeting of 1h?