: It uses built-in generative AI models to "listen" to dialogue in real-time, automatically generating accurate captions even for videos that never had them.