Create Cinematic AI Videos with MiniMax H3

Use one clear, high-quality reference image and state exactly what must remain unchanged: facial identity, hairstyle, clothing, colors and accessories. Keep the character visible at a reasonable size, avoid several major pose changes in one short clip and do not use competing character references. For example: “Preserve the exact face, short black hair, red jacket and silver necklace from Image 1 throughout the continuous shot.

Start with a sharp product image that shows the geometry and branding clearly. Ask H3 to preserve the exact shape, proportions, materials, colors and logo, then request camera movement around the product rather than a transformation of the product itself. Put required on-screen wording inside quotation marks and avoid hiding the label behind glare, hands or extreme motion.

Replace a broad request such as “make the person move” with one achievable action described in sequence. Define the starting pose, weight shift, main movement and ending pose, and keep the camera direction simple. If available, add a short reference video that clearly demonstrates the motion and tell H3 to use it only for movement and timing.

Describe movement at three levels: subject motion, environmental motion and camera motion. For example, ask the subject to turn toward the window, curtains to move in the breeze and the camera to track slowly to the right. A specific motion plan gives H3 more useful direction than “animate this image” or “make it cinematic.

The prompt may contain too many actions, conflicting camera instructions or unclear priorities. Put the essential subject and action first, write events in chronological order and use one primary camera movement. If several references are included, assign one explicit role to each instead of expecting H3 to decide which details matter most.