h3-two-part-character-reveal

H3 Two-Part Character Reveal

Design both halves together so the second clip begins from a declared handoff state and expands the visual language instead of repeating it.

Route the Request

• For prompt writing or refinement, return text only.
• Read references/style-blueprint.md and use $h3-prompt-writing.
• Invoke $h3-video-generation only when the user explicitly names the part to render. Never submit both independent jobs implicitly.
• Use native Ref2VA for multi-image identity, pose-library, or reference-video motion. Audio embedded in a reference video is forwarded automatically; standalone reference-audio files are not exposed by the current CLI contract.

Assign Reference Authority

1. Picture 1 is the complete pose-and-world library: body, costume, hair, accessories, environment, lighting, and approved poses.
2. Picture 2 is face identity only. It cannot replace costume, body, environment, or pose logic from Picture 1.
3. Supplementary pose images may add approved poses only when they preserve the same character and scene.
4. A motion reference controls timing and movement quality, not identity or costume. A BGM reference controls energy structure, not copyrighted melody reproduction.

Write the Two Parts

• Give each 15-second half its own complete prompt and shot ledger. The full pattern may use about 13 concise cuts per half when every cut has one purpose.
• Part 1 establishes place, materials, silhouette, hands, accessories, eyes, and full identity. It ends on a clearly described pause-state handoff.
• Part 2 begins from that exact declared camera, pose, expression, lighting, ambience, and musical energy. It escalates to action peak, hero framing, and a final key visual.
• Repeat the seam contract verbatim in both prompts: same character, costume, environment, final/opening pose, light direction, ambient reverb, and music energy.
• Give Part 2 a non-repetition list naming every Part 1 shot mechanism it may not reuse.
• Use only approved poses. Mark which poses Part 1 consumes and reserve the remaining poses for Part 2.
• Allocate detail priority as face > eyes > hands > environment whenever motion and resolution compete.