In my experience, transcription software has no problem with transcribing sped up audio, or audio that is inaudible to humans or extremely loud (as long as not clipped), I wonder if LLM transcription works the same.