MiniMax H3 reference-to-video is a video generation model that uses your photos as a reference for who appears and a text prompt for what happens. Give it one to four photos, refer to them as Image 1 to Image 4, and describe the scene, the action and the camera. It returns a new 5 to 15 second clip with a generated soundtrack. Nothing is taken from an existing video: the place, motion and sound are all new. fal lists the model at $0.05 per second at 480p and $0.06 at 768p; here a second costs 10 or 12 credits.
New to this? Learn how turning a photo into a video with AI works.
