minimax-h3-prompt-standardizer

Provide your one-line idea and the Skill expands it into a full H3 prompt covering subject, scene, timeline, camera, lighting, and sound. It infers sensible defaults for duration and aspect ratio, asking at most one question only when a missing detail would change the result.

The Skill segments stories at visible boundaries such as shot size, location, or completed actions, never mid-pose or mid-emotion. Each segment gets its own prompt with a tail-frame state and cut point to maintain continuity across generations.

Yes, the Skill defines an upload order labeling each reference image's role, such as character identity or scene lighting. It designates a primary reference when images conflict and instructs the model to ignore unwanted backgrounds, watermarks, and artifacts.

No, it only produces text prompts to paste into MiniMax H3. It does not generate, render, or edit video files, and it never claims to see assets that were not provided.

Failures come from overloaded actions, vague wording, or missing continuity anchors. The Skill reduces simultaneous actions, replaces vague terms with concrete visual descriptions, and adds targeted constraints against identity drift, lip-sync errors, and camera jumps.