OpenAI's open-source automatic speech recognition (ASR) system that approaches human-level robustness and accuracy on English speech recognition.