This ComfyUI tutorial template shows how to turn a single image into an animated video using MiniMax H3 Max via the MinimaxHailuo03ReferenceNode. The core path is simple: you load a still frame with LoadImage, describe the motion in PrimitiveStringMultiline, and the MinimaxHailuo03ReferenceNode generates a video while preserving the source composition and aspect ratio. SaveVideo writes the final clip to disk, and SaveText can archive the exact prompt and response metadata for reproducibility.

Two optional paths are controlled by red ComfySwitchNode gates. The Prompt Enhancer (Optional) stage lets you route an alternate or refined prompt into the generator if you want cleaner wording without changing your base prompt field. The Upscale Video to 2K (Optional) stage adds a post-process branch for higher-resolution output, but it only accepts a 768p source—so you must generate at 768p in the core node if you plan to upscale. For more advanced control, MinimaxHailuo03ContextIRNode can be added to supply context/image-reference signals for stronger identity retention, and MinimaxHailuo03RegenerateNode enables quick re-runs with new seeds without reconfiguring inputs. This makes the template practical for animating character portraits, product shots, and concept visuals with prompt-driven motion while keeping the original look intact.