Hailuo 2.3-Fast Image-to-Video: Single-Frame-Driven Dance and Product Shots

This entry is aimed at developers who need to quickly produce dynamic assets from a single first-frame image, demonstrating the real input structure of Hailuo 2.3-Fast (MiniMax, image-to-video) on the author's pipeline. The author's schema requires only two fields, prompt and image_url, and additionally offers two optional control slots: duration (6/10 seconds) and prompt_optimizer; it does not support first and last frames, and can only extend forward from a single first frame, so planning should focus on "an extendable first-frame composition and a clear direction of motion." The work uses an anime-style full-body dance shot as the main pipeline to test the stability of complex full-body motion and artistic style: the first frame provides a clear character silhouette and negative space, and the prompt describes holding a medium shot and a slow rightward camera move, so the direction of motion matches the first-frame composition. Only non-media parameters such as duration and prompt_optimizer are placed in base; the first-frame image is generated by the pipeline and then bound to image_url via bind.path, ensuring the submission snapshot matches the real output. Use cases: quick motion test shoots based on an existing character reference or concept art; using short English prompts to verify how the camera-move description works with the first frame; iterating on character performance shots at lower cost before finalizing. Note that descriptions of resolution and speed in the retrieved materials come from third-party pages; this entry makes no parameter promises and only presents the control items already declared in the schema.