Deliverables: style lock + (optional) storyboard table + single-image/series-image/video H3 prompt. Default model intent: MiniMax-H3.
Global lock (repeated in every prompt)
- handmade paper-craft stop-motion, miniature theater box, multi-layer cardstockvisible thickness
- fibers/creases/torn edges,real shadows between layers, macro depth of field
- paper puppet separate parts; label arrows are also paper pieces
- Motion: small segmented movements, brief pauses, slight rebound, hinges/pull-tabs/sliding rails
- Forbidden: silky CG, plastic 3D, flat vector without paper edges, realistic large motion
Before writing
Theme · one-sentence learning objective · audience · duration (15/30/60) · aspect ratio (default 16:9) · deliverable: single image / series / 5s video / storyboard+video
If you only need a single item: just write the corresponding prompt, no need to force the full package.
Creative direction (choose 1 of 3 to write into the style sentence)
Pop-up book journey / paper scientist laboratory / layered cross-section / miniature nature theater / paper mechanism board
Storyboard table (serves shot writing)
| Time | Knowledge beat | Visual | Paper-craft motion | Camera movement/transition | Sound effect intent |
|---|
15s≈4 shots; 30s≈5–6; 60s≈7–9;one knowledge point per shot。
Camera movement: slow push-in, lateral parallax, fixed explainer, macro texture.
Transitions: page turn, pop-up unfold, pull-out tab, paper cloud—no electronic glitch neon (unless the user wants it).
Prompt type
Single image: theme words + global lock + one layer of main knowledge + macro paper sculpture
Series images: shared material, light direction, identity; each image advances only one step
5s image-to-video: preserve paper texture and layering; slow push-in/parallax; mechanisms move slightly; no melting, high-speed orbiting
Segmented video (≤15s/segment): per-second paper motion + inherited character, set, light, and color lock + native audio: paper-touch SFX + lightly educational BGM (can be N/A)
[负向常挂] plastic CG, photoreal skin, flat vector no paper edge,
no layer shadow, liquid morph, cyber neon glitch, fast 360 orbit
When field-based, connect h3-prompt-writing。
Self-check
- paper thickness and interlayer shadows are visible
- one knowledge point per shot
- motion looks like stop-motion paddle animation
- memory sentence/learning objective lands in the final shot