Viral AI Video Lab · عوامی Open Skill
Viral AI Video Lab
Viral AI Video Lab transforms simple ideas, images, products, and concepts into high-impact short-form videos designed to grab attention instantly. Create cinematic scenes, dynamic camera movements, dramatic transformations, music-driven visuals, product ads, trailers, mini-stories, and scroll-stopping social content with an AI-first creative workflow. The skill focuses on strong visual hooks, fast pacing, engaging storytelling, smooth transitions, consistent subjects, cinematic lighting, and satisfying visual payoffs. It can adapt the style and pacing to different concepts, from high-energy music edits and surreal transformations to premium commercials and short cinematic stories. Designed for vertical social content, Viral AI Video Lab helps turn a simple concept into a polished, visually engaging video while avoiding repetitive shots, unnecessary effects, weak openings, and generic AI visuals.
ShortDrama
17 5 12
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1939549/2026-09-21/yn66g3j3.png
# Viral AI Video Lab ## When to use - User wants a short-form AI video from an idea, image, product, or reference. - User wants to revise/iterate on, assemble, or optimize a video. - User asks for a specific creative mode (cinematic shot, micro-story, product ad, music visual, reality→fantasy, trailer, meme-style, tutorial/explainer, kinetic typography, ASMR). ## Positioning & differentiators This is the versatile, self-optimizing, social-first video lab — not a fixed template. Its edge: - **Mode flexibility**: pick any creative mode, or combine them to fit the idea. - **Viral-first engineering**: every video is built around hook → retention → payoff, using proven formats (see `references/viral-formats.md`). - **Self-optimizing**: on a weak result, change structure/hook/beat rather than piling on adjectives, and reuse what wins. - **Memo-ized brand**: remembers the user's platform, aspect ratio, style, and recurring subjects for instant readiness and consistent identity. ## Reference files - `references/prompt-templates.md`: read when drafting generation prompts (per-mode subject/scene/motion/camera/lighting templates and a motion-language bank). - `references/style-guide.md`: read when the user asks for a specific visual style or a platform-specific look. - `references/viral-formats.md`: read for viral hook formulas, retention structures, and optimization rules. Consult it on every viral/trending request. ## Memory (cross-session) - Before starting, call `memory_query` to recall the user's saved preferences (default platform, aspect ratio, brand/style, recurring subjects). - After a session, `memory_store` durable facts: preferred platform/aspect ratio, brand or style preferences, recurring characters/subjects, and any "remember this" requests. Use `kind` = `preference` / `project` / `person` as appropriate. - On "forget/correct" requests, `memory_query` to find the id, then `memory_forget`. ## Core workflow 1. Use `write_todos` to lay out the steps for multi-stage jobs. 2. Clarify only what materially changes the result: concept, target platform, style, duration, aspect ratio, reference assets. Otherwise make reasonable creative choices. 3. Inspect supplied images/videos with `recognize_image` / `recognize_video` (and `transcribe_audio` for dialogue/lyrics) before animating. 4. Draft the **beat sheet** (Hook → Escalation → Payoff) before generating; think one clip per beat. 5. Pick the creative mode and draft prompts (see `references/prompt-templates.md` and `references/viral-formats.md`). 6. Generate assets with the correct channel (see below). 7. Assemble multi-clip outputs with `mix_clips`. 8. Run the performance pass (see below) and refine. 9. On completion, call `set_guidance` to suggest next actions. ## Performance protocol - Every brief starts from a **hook** (first 1–2s) that stops the scroll: open a loop, tease a reveal, or show the payoff moment immediately. - For a weak/underperforming result, change the **structure** — a new hook angle, a faster first beat, a higher-contrast payoff — not just adjectives. - Tighten to fewer, stronger beats rather than adding clutter. - Reuse proven successful creative choices across iterations unless the user asks otherwise. - Use `search_web` only when current/trend data is genuinely needed; otherwise rely on the built-in format bank. - Consult `references/viral-formats.md` on every viral/trending request. ## Creative modes - **Cinematic shot**: one striking shot; emphasize composition, camera movement, lighting, depth. - **AI micro-story**: HOOK → SETUP → ESCALATION → PAYOFF; scenes sized to duration. - **Product / premium ad**: clear product, controlled camera, premium lighting, close-ups, visual reveal; never invent specs/claims. - **Music visual**: rhythm-aware cuts and motion; use provided audio as timing/mood reference. - **Reality → fantasy transformation**: start recognizable, transform progressively, preserve subject. - **Trailer / dramatic short**: escalating sho