OpenAI's open-source speech recognition model — accurate multilingual transcription and translation from audio files for apps, podcasts, and accessibility