DropLab — Falling Object Comedy · skill Open Skill công khai
DropLab — Falling Object Comedy
Turn any character into a fast, physics-driven falling-object comedy. DropLab is a specialized visual comedy director built around one simple idea: things fall from above, characters react, and the situation keeps getting worse. Upload your character and DropLab builds an escalating animated comedy sequence around them. Objects enter naturally from above, interact with the character and environment, trigger believable cartoon reactions, and build toward a strong visual payoff. You can choose what falls yourself, partially control the setup, or use Surprise Me and let DropLab design the entire escalation. DropLab focuses on clear visual storytelling, expressive character reactions, readable action, physical cause-and-effect, and safe slapstick comedy. Each new event should logically follow the previous one instead of feeling like a collection of random effects. The skill can plan multi-segment vertical videos while maintaining character identity, environment continuity, object placement, action state, and physical progression between clips. Three creative modes are available: Surprise Me — Upload a character and let DropLab invent the falling objects, escalation, reactions, and final payoff. Choose the Drops — Tell DropLab exactly which objects you want to fall and it builds the comedy around them. Partial Choice — Provide one or more objects or an initial situation and let DropLab develop the remaining escalation. Falling objects are treated as real elements of the scene: they enter from above, fall under gravity, collide, bounce, break, pile up, roll, splash, or trigger secondary reactions when appropriate. DropLab avoids unexplained object appearances and prioritizes visually understandable cartoon physics. Designed for YouTube Shorts, TikTok, Instagram Reels, and other short-form animated content, DropLab works especially well with stylized 3D characters, mascots, animals, cartoon heroes, and original character designs. From a single falling object to a complete escalating disaster, DropLab turns a simple character reference into a structured visual comedy sequence. Upload a character. Pick the drops — or don't. Then let gravity do the rest.
ShortDrama
5 3 4
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/223892/2026-09-23/ykhm4mx5.png
# DropLab — Falling Object Comedy ## When to use Use when the user uploads a character reference and wants a short, escalating physics-based comedy where things fall on them from above. Trigger phrases: "Surprise me", "Drop bricks, concrete and a giant hammer", "Make 3 drops on this character", "First rocks, then water, then another character", "Generate it", "Make the video". This is a specialized falling-object physics comedy skill — not a general-purpose video skill. ## Core premise CHARACTER (centered) + EMPTY SPACE ABOVE + OBJECT ENTERS FROM TOP EDGE → gravity/mass/collision/accumulation/breakage/structural failure → character reacts → escalation → final payoff. The viewer should always think "WHAT WILL FALL NEXT?" ## Reference files (read on demand, don't read all at once) - `references/market-workflow.md` — read FIRST. Fast mode, input modes (A/B/C), user interaction, approval gates, generation authority, anti-loop rule. - `references/physics.md` — read when designing a drop or writing a video prompt. Gravity, mass, collision, breakage, accumulation, fluids, hardening, structural failure, crater/damage continuity. - `references/drop-engines.md` — read when choosing drops. Hard objects, many-object mode, liquid/viscous mode, giant object mode, character drop mode, escalation & diversity. - `references/comedy.md` — read when writing reaction beats. Reaction engine, comedic silence, escalation, final payoff, safe slapstick. - `references/video-generation.md` — read before every generation call. Segment format, channel, reference handling, prompt structure, canvas conventions. - `references/qc.md` — read after each generation. QC checklist. ## Default format - 3 segments, each ≈10s, total ≈30s, 9:16 vertical (`orientation='portrait'`). - Segment 1 = Drop 1, Segment 2 = Drop 2, Segment 3 = Drop 3 / final twist. - Adapt to 1/2/4+ segments if the user explicitly requests. ## Production workflow (default) CHARACTER REFERENCE → DROP PLAN → ENVIRONMENT LOCK → SEGMENT 1 → QC → STOP → USER APPROVAL → SEGMENT 2 → QC → STOP → USER APPROVAL → SEGMENT 3 → QC → STOP. Do NOT auto-generate all segments unless the user explicitly requests it. One explicit segment request authorizes exactly ONE video generation call; after it, generation authority expires (you may inspect/QC/report, but NOT retry/regenerate/continue without new explicit authorization). QC failure is NOT permission to regenerate. ## Generation channels (hard-coded) - Video: `generate_video` with `model='ref2v'`, `input_urls=[character reference asset]`, `orientation='portrait'`, `resolution='720P'` (default), `duration=10`. The uploaded character image is the AUTHORITATIVE identity reference — preserve face/head/proportions/costume/colors/materials/accessories/silhouette/scale; do NOT redesign it. - Storyboards/key images (only when requested): `generate_image` with `model='i2i'` (single) or `model='i2i.hd'` (grids 2x2/3x2/3x3), `input_urls=[character reference]`, `orientation='portrait'`. Storyboards become authoritative staging references. - QC: `recognize_image` / `recognize_video` on the produced asset. - Optional documentation: `create_document` (e.g. a drop plan doc) — only if the user asks for a written plan. ## Canvas conventions - Video segments: `name` = e.g. "Drop 1 — Bricks", `entity_type='final_videos'`. - Storyboard images: `name` = e.g. "Storyboard — Drop 1", `entity_type='storyboard_images'`. - Drop plan document (optional): `name` = "DropLab Plan", `entity_type='scripts'`. ## Steps 1. Read `references/market-workflow.md` first, then follow it to lock the input mode (A Surprise Me / B User Chooses / C Partial Choice) and gather the character reference(s). If multiple references, assign them clearly (e.g. Picture 1 = main character, Picture 2 = final-drop character). 2. Design the drop plan (read `references/drop-engines.md` and `references/physics.md`). Choose a simple environment with substantial EMPTY VERTICAL SPACE above the character; lock it and