From Text to Image: MiniMax H3 Max First-and-Last-Frame Shot Transition

Using the storyboard concept of "text → image" as an example, this demonstrates how to use first-and-last-frame control in the MiniMax H3 Max image-to-video pipeline: two assets are used to generate the two key frame images for the storyboard's opening and closing respectively, and then bound via first_frame_image and last_frame_image so that the model completes the camera movement and action transition between the given opening and closing compositions. Structure of this piece: the first frame is the storyboard's starting point (character seen from behind, empty street, cool-toned atmosphere), and the last frame is the closing image of the same scene (character turns around, streetlights come on, warm-toned atmosphere). The prompt describes only the camera language, action rhythm, and environmental changes, without restating the static composition—the compositional information is carried by the two reference frames. This is suited to rapid multi-shot storyboard prototyping for short videos and ads, shot-transition validation for turning illustrations into animation, and shot iteration scenarios that require before-and-after comparison. On the parameter side, the pipeline defaults are used: 768P resolution, 5-second duration, and balanced prompt-expansion mode; the first frame is required while the last frame is optional, and this piece provides both to achieve a controllable transition.