According to the GitHub repo this is an "image-to-video model". They tease of an upcoming "text to video" interface on the linked landing page, though. My gue

According to the GitHub repo this is an "image-to-video model". They tease of an upcoming "text to video" interface on the linked landing page, though. My guess is that interface will use a text-to-image model and then feed that into the image-to-video model.