Generating 3D Science Viral Shorts
Given a topic (e.g., "How vaccines work" or "Why cats always land on their feet"), output a full scene-by-scene video blueprint — not just a script summary. The user should be able to hand the output directly to a 3D animator or AI video generator.
Example trigger: "Make me a short about how stomach acid digests food"
→ Respond with the full blueprint structure defined below (Concept Title & Hook, Scene Breakdown, AI Prompts), not a prose explanation of digestion.
Progress:
- Step 1: Identify the scientific/physical concept to explain
- Step 2: Invent an absurd analogy using humans, animals, or objects as stand-ins for the process
- Step 3: Write the Hook (0-3s)
- Step 4: Break down the Core Process (3-20s) into 3-6 scenes
- Step 5: Write the Payoff/Punchline (20-30s+)
- Step 6: Fill in scene table with Visual/Action, On-Screen Text, Voiceover, Sound Design
- Step 7: Generate AI image/video prompts for 2-4 key frames
- Step 8: Sanity-check pacing (no dead air) and scientific accuracy under the humor
Step 1-2 (Concept → Analogy): Pick a literal physical stand-in. Scale matters: shrink humans/animals to cellular scale, or scale up objects to represent molecules. The analogy must map 1:1 onto each stage of the real process (don't just decorate — structurally mirror cause and effect).
Step 3 (Hook): Must establish the absurd premise visually within the first shot, before any explanation. Lead with the weird image, not a setup sentence.
Step 4 (Core Process): Each scene = one discrete step of the real-world process. Alternate between "macro" (character-scale action) and "micro" (molecular/cross-section close-up) shots to keep visual rhythm.
Step 5 (Payoff): Resolve the analogy's story AND land the scientific point in the same beat — the humor should come from the audience realizing both happened at once.
Hook: [1-2 sentence description of opening visual + hook line]
Scene Breakdown
| Timestamp | Visual & Character Action | On-Screen Text | Voiceover Script | Sound Design |
|---|---|---|---|---|
| 0-3s | ... | ... | ... | ... |
| 3-8s | ... | ... | ... | ... |
| ... | ... | ... | ... | ... |
AI Generation Prompts
Frame 1 (Hook shot): [detailed prompt: subject, style, lighting, camera angle, color palette, 9:16 framing] Frame 2 (Key process shot): [detailed prompt] Frame 3 (Punchline shot): [detailed prompt]
Every AI prompt must include: stylized 3D clay-render aesthetic, vertical 9:16 framing, bright saturated contrasting colors, macro/micro lens detail where relevant, and explicit camera angle.
Example 1: Input: "Explain osmosis" Output:
Hook: Two tiny soldiers made of water stand on opposite sides of a glowing cell-wall fence, one side packed shoulder-to-shoulder, the other nearly empty — text flashes "WHY YOUR CELLS ARE PEER-PRESSURED 24/7"
Scene Breakdown
| Timestamp | Visual & Character Action | On-Screen Text | Voiceover Script | Sound Design |
|---|---|---|---|---|
| 0-3s | Crowded blue water-soldiers crammed behind a semi-permeable fence, bouncing off each other; empty yard on the other side | "TOO MANY ROOMMATES" | "Meet the most crowded apartment in biology." | squeaky rubber bounces, crowd murmur |
| 3-10s | Soldiers squeeze one by one through fence gaps to the empty yard, fence shown in cross-section glowing with tiny gate animations | "SQUEEZING THROUGH THE FENCE" | "Water always slides toward the emptier room — no invitation needed." | soft 'pop' per soldier passing, whoosh |
| 10-20s | Both yards now evenly spaced soldiers, fence stops glowing, soldiers high-five | "BALANCE ACHIEVED" | "Once it's even on both sides, the moving stops. That's osmosis." | satisfying chime, crowd sigh of relief |
| 20-28s | Camera zooms out: the whole yard was inside a single cucumber slice going limp in a salt bowl | "...AND THAT'S WHY YOUR CUCUMBER WENT SAD" | "Same reason your pickles shrink. Science: it's just everyone wanting more space." | comedic slide-whistle, crunch |
AI Generation Prompts
Frame 1 (Hook shot): Stylized 3D clay-render water-droplet soldiers, bright cobalt blue and white, crammed together behind a glowing translucent cell-membrane fence, empty golden-lit yard on the right, vertical 9:16 framing, soft studio lighting, wide shot slightly low angle, vibrant saturated colors, Pixar-like character design. Frame 2 (Key process shot): Macro cross-section view of a glowing cell membrane with small round gates, blue water-soldier characters squeezing through one gap mid-motion, motion blur droplets, vertical 9:16, bright rim lighting, clay-render texture, shallow depth of field. Frame 3 (Punchline shot): Wide pull-back 3D render revealing the entire soldier-yard scene is inside a translucent cucumber slice sitting in a salt-filled glass bowl, dramatic zoom-out camera move, bright saturated green and white palette, vertical 9:16, cinematic macro-to-wide transition.
- Always pick analogies with clean cause-effect mapping — avoid decorative metaphors that don't actually teach the mechanism.
- Keep voiceover lines short and quotable; they double as the on-screen text source.
- Specify "no background music" in sound design — foley/ASMR only, per style constraints.
- Reuse consistent character/color designs across scenes for viewer recognition (e.g., always blue = water, red = heat/energy).
- End on a punchline that reveals the real-world scale/context (zoom out to reveal the mundane object the analogy was secretly happening inside).
- Don't front-load explanation before the hook — lead with the weird visual first.
- Don't let any scene exceed ~8 seconds; shorts die from pacing lag, not lack of detail.
- Don't sacrifice scientific accuracy for the joke — the punchline must still teach the real mechanism.
- Don't mix horizontal/cinematic framing language into prompts — always reinforce 9:16 vertical explicitly.
- Don't forget macro/micro alternation — an entire video at one scale feels visually flat.