“Black Forest Labs is developing multimodal models trained on images, video, and audio that also incorporate action prediction, enabling them to be deployed on robots.”