Creates videos, images and text from multimodal inputs with up to 20 reference assets. Supports native 30s video generation.
