Melia, our new multilingual speech model, can transcribe an hour of audio in under 20 seconds. ⚡ For teams processing audio at scale, that means analysis, search and downstream AI can start almost immediately. For James, it means a very short LEGO challenge. ... and as a