I developed my own solutions, pretty rudimentary - it divides the MP3s into chunks that Whisper is able to handle and then sends them one by one to the API to transcribe. Works as expected so far, it's just a couple of lines of Python code.