The most controllable AI product video starts from a real product image. Wan image-to-video can add a camera orbit, moving light, subtle environment motion, or a reveal while the source image supplies the composition.
Use a clean high-resolution image, clear separation from the background, enough crop space for camera movement, and a label that is already correct. AI video can distort typography or geometry, so every final frame requires review.
The camera makes a slow 20-degree clockwise orbit. A narrow warm highlight travels left to right across the product. Preserve product silhouette, material, label geometry, color, and background composition. No new text, no extra objects.
Pause on the first, middle, and last frame. Check logo spelling, edges, reflections, object count, surface texture, and whether the product changes size or shape. A smooth video can still contain an unusable middle frame.
It can preserve source structure, but exact typography is not guaranteed. Keep motion simple, state the preservation constraint, and inspect every frame before commercial use.
Use an image when product identity matters. Text-to-video is better for concept exploration where exact shape and branding are not required.
The current planned low-cost image tier is Wan 2.2 at 15 credits. It appears in the selector only after a paid canary verifies the endpoint and playable output.