What problem does it solve? Writing prompts for the MiniMax H3 video model fails when everything is crammed into one paragraph, reference files lack declared roles, or audio instructions contradict each other. This Skill applies the official MiniMax prompt formula to the local ComfyUI video_minimax_h3_i2v workflow so generations match the intended scene, timing, and sound. ## Core Features & Use Cases - Official three-block formula: structures prompts as reference material notes, core idea, and scene-by-scene timeline with [0s-1.5s] style marks matched to the configured duration. - Common mistake prevention: covers undeclared file roles, contradictory music instructions, vague camera moves, and single-take vs. multi-shot narration errors. - Local workflow procedure: details editing the subgraph instance node 105 widgets_values prompt with python_embeded, uploading images via upload_file, validating, running, and verifying output MP4s. - Use Case: Given a product photo and a description, produce a 5-second cinematic commercial clip by writing a template-based prompt, syncing it into the workflow, and polling the job until the video renders. ## Quick Start Write a MiniMax H3 prompt for a 5-second cinematic video using my image at input/product.png and run it through the local i2v workflow.
