Let's try the Whisper large-v2 model from OpenAI.