Viral Hook

pika-labs/pika-plugins/skills/viral-hook

作者 pika-labsf27b3ba28a7b无许可证40 个星标收录于 2026年10月9日更新于 2026年10月8日仓库2个月前更新

Use when the user asks for viral hook or a task matching the examples below. Prepend a 4s viral hook + optional designed title to a user's video, then hard-cut into the real clip. An attention-grabbing event erupts into the user's OWN scene; the title is rendered in-scene by Seedance. One MCP call does the whole render. Requires Pika MCP. Triggers: "add a viral hook", "disruption hook on my video", "hook + title on my clip", "viral hook with typography", "scroll-stopper intro".

AI 生成的概览

通过 Pika MCP 渲染调用,为用户视频前置一段吸睛钩子和可选标题。

功能
该技能引导智能体在用户视频开头加入一段约四秒、无对白的极端钩子,然后硬切回原始片段。它会读取场景,从破坏性闯入、物理异常、内容放大或内容抛射等类别中选择钩子动作,并可选地撰写一句全大写短标题及排版方向。一次 generate_viral_hook 调用即可完成渲染,并同时返回钩子片段和拼接后的成片。
适用场景
当用户要求为自己的视频添加病毒式钩子、抓眼开场、干扰式钩子或带文字的钩子时使用。适用于希望在真实画面之前加入吸睛开头的短视频内容。
运行要求
需要具备 upload_asset、analyze_media、generate_viral_hook 和 task_status 能力的 Pika MCP,以及视频文件或 https 链接。该技能不附带脚本,仅为说明文档。

Viral Hook

Prepends an extreme, no-dialogue ~4s hook to a user's video — an attention-grabbing event that erupts into the user's own scene, with an optional designed title burned into the lower third — then hard-cuts into the untouched clip.

generate_viral_hook does the whole render in one call and returns BOTH the hook and the stitched final. This skill's job is the creative judgment the tool can't make: read the scene, pick a hook action from the menu, and write a title.

Hook rules (govern what you WRITE as hook_action)

  1. Certain to grab attention, fast. Extreme and unmissable inside ~2s — over plausibility or relevance to the video's topic.
  2. Erupts into the user's OWN scene. Same person, location, lighting — the tool anchors on the best visual anchor frame: 0s when usable, otherwise the first detectable face frame. Describe the event entering THAT space, never a different setting.
  3. Use the scene's real geometry. Enter through a doorway with depth, a wall behind the subject, headroom above.
  4. No dialogue. Voice-free (only ambient / impact SFX) — don't write spoken lines into hook_action.

Hook super-category menu

Four super-categories. Rotate super-category per run on the same clip/session — without rotation, regenerates collapse to vehicle / explosion / creature. Archetypes are starting points, not a closed list; invent freely within the category.

A. Destructive intrusion — external force violently enters and damages the scene

ArchetypeWhat happensScene affordance
Incoming vehiclecar / truck / bus smashes through a wall or doorway and barrels at camerabackground depth or a back wall
Flash flood / wavea wall bursts and water surges down toward cameracorridor / back wall
Structural collapseceiling / wall / shelf caves in, debris rains downheadroom / walls
Explosion / blastfireball erupts behind them, shockwave + debris rush forwarda background surface
Creature attackanimal / monster lunges from off-frame or bursts inopen off-frame edge
Absurd intruderdinosaur, elephant, UFO, giant hand enters behind thembackground space
Natural disastertornado / meteor / lightning / sinkholeexterior or large space

B. Physics anomaly — laws of physics break, photoreal; scene + subject stay, motion goes wrong

ArchetypeWhat happensScene affordance
Gravity flips upwardsubject, loose objects, and dust fall UP toward the ceilingany indoor scene
Time freeze around subjectevery other object holds mid-air while the subject keeps movingscene with loose / moving objects
Magnetic pull from ceilinghair, clothing, jewelry, small objects yank upward as if a ceiling magnet firedindoor scene + loose props
Room rotates 90°walls + floor rotate around the upright subject until "down" is now a wallenclosed space, visible walls + floor
Object orbitall small loose objects lift off and orbit the subject like a tornadoscene with props on surfaces
Inverted color flashthe scene briefly inverts to negative colors and back, like a glitchany scene

C. Content-supersized — an element FROM the clip appears building-sized and crashes in

Pull the element from the scene reading (Step 2) — hero product, pet, drink, phone, logo.

ArchetypeWhat happensSource element
Giant hero productthe product being shown appears at 10× scale and smashes through the wall / ceilingthe product in the clip
Building-scale petthe subject's pet appears at building height in the window / doorwaya pet in the clip
Giant drinkthe cup / can / bottle becomes building-sized and tips over the rooma drink in the clip
Giant phonethe subject's phone scales up to fill the back wall, screen glowinga phone in the clip
Giant logoa brand mark on a wall / shirt / packaging grows to cover the back wallany logo in the clip

D. Content-projectile — a relevant object FROM the clip flies at the subject at speed

Source from the scene reading.

ArchetypeWhat happensSource element
Hero-object slamthe product they're showing flies at high speed into their facethe hero product
Food / drink splashtheir food or drink hurtles into them, splattering everywherefood or drink in the clip
Swarm peltmany copies of a small object (sneakers / pills / cans) pelt them like haila small repeating object
Prop whipan object on a surface (mug, lamp, book) whips off and slams into themany heavy prop visible

Title rules (when the user wants one)

  • ≤7 words, ALL CAPS, one punchy line. Tie it to BOTH the hook action AND the clip's actual content. Example: giant-sneaker projectile + sneaker unboxing → "THIS DROP HITS DIFFERENT."
  • Optional — omit for a title-free hook.
  • Typography defaults to bold distressed display type with a thick black outline. The user may override via type_style (a free-form style string) or by pasting a brand.md — extract ONLY the typography directives (fonts, weights, treatments), ignore logos / palette / voice. Warn that thin / un-outlined styles may wobble over the 4s.

Workflow

Step 0 — Intake (empty args)

If no video was provided, print this menu verbatim and STOP — do not call any tool:

What would you like a viral hook on? Paste any of:

  • A local video path — e.g. /Users/me/Desktop/clip.mp4
  • An https URL to an mp4
  • A path + a title — e.g. clip.mp4 — "YOUR PROMPTS WON'T SAVE YOU"
  • A path + "surprise me" — agent picks the hook action and title

Wait for the next message. Don't guess an input.

Step 1 — Get a URL

If the input is a local file, upload_asset it and use the returned public_url. If it's already an https URL, use it directly. Any format is fine — no transcoding or probing here.

Step 2 — Read the scene

analyze_media(video_url) → capture the subject + framing, indoor/outdoor + setting, lighting, photoreal vs stylized aesthetic, and the spirit/topic. Also note the hero objects (product, pet, drink, phone, logos), including whether each one is visible near the opening or only appears later — required for super-categories C and D. This reading is passed verbatim as the scene parameter.

Step 3 — Pick the hook action

Rotate super-category vs the previous run on this clip; pick a different one if possible. Choose the strongest archetype the scene's geometry supports (A/B) or the most prominent hero object (C/D). Write the specific hook action — entry vector + motion + scale + peak — as one vivid sentence. This is hook_action. For C/D, name the exact hero object so the tool ties it to the clip. If the hero object appears later in the clip rather than near the opening, make the first hook frame explicit: "from the first frame, the subject is holding/using [hero object]".

Step 4 — Title + style

Ask whether the user wants a title. If yes, propose a ≤7-word ALL-CAPS line tying the hook action to the topic, get approval, and offer typography choices before rendering: default, custom direction, or brand.md. If the user has no preference, use default and say so. If no title is wanted, skip — the hook renders title-free.

Step 5 — Render

Call generate_viral_hook({ video_url, scene, hook_action, title?, type_style? }). If it returns {task_id, status}, poll task_status(task_id) in a tight loop until terminal, then read the result. On failed, surface the error — don't retry blindly.

Step 6 — Deliver

Present both clips:

  • hook only: [[video:<hook_url>]]
  • final stitched: [[video:<final_url>]]

What NOT to do

  • Don't re-implement the render in the skill. No extract_frame / generate_image / edit_concat / transcode here — generate_viral_hook owns all of it. If the hook looks wrong, change the hook_action / title text, not the pipeline.
  • Don't pass a seed. Variations should come from hook_action, title, or type_style.
  • Don't bake a title onto the real video. The title belongs only on the hook.
  • Don't auto-edit an approved title. If a rephrase is needed, ask.

来源与署名

来源:pika-labs/pika-plugins位于skills/viral-hook提交f27b3ba

许可证: 无许可证

内容归原作者所有。SourceWeft 从公开仓库中收录这些内容。

举报或申请下架

更多来自 pika-labs/pika-plugins 的技能

Voxel It

pika-labs

把照片变成 Minecraft 风格画面,人物保持写实,仅将物体、背景或两者体素化。

Design & Creative402个月前更新

Vfx

pika-labs

Use when the user asks for vfx or a task matching the examples below. Turn a plain video clip into a cinematic AI-VFX shot at 1080p by default (or 4K / 720p on request). Give Claude a video and the change you want; it reads EVERY frame via local contact sheets and understands the audio, writes a Seedance-faithful prompt that locks your face, gestures, and camera move, then re-renders the same shot with the VFX baked in via Seedance reference-to-video.

待分类402个月前更新

Ugc Ads

pika-labs

Use when the user asks for ugc ads or a task matching the examples below. Multi-cut jump-cut UGC product ad — HOOK + 3 JUMP CUTs + OUTRO, 15s, 9:16 vertical (3:4 optional, seedance only), POV first-person talking-head selfie, every beat has spoken dialogue with native lip-sync, 5-act narrative arc (set → name → reveal → twist → punchline). Six category essences (HAUL / APP / FOOD / BEAUTY / FITNESS / TECH) auto-picked from the input URL. Creator-style raw UGC talking-head with multi-beat conversational dialogue. Use when the user asks to "make a UGC ad", "jump-cut product ad", "POV product reveal", "creator-style ad", "haul-style ad", "unboxing ad", "TikTok-style product video", or "talking-head ad about [URL]".

待分类402个月前更新

Stagefight

pika-labs

生成观众手机拍摄的舞台化 cosplay 对战短片:先用 gpt-image-2 生成舞台静帧,再生成约 10 秒 16:9 的图生视频对打片段。

Design & Creative402个月前更新

Podcast

pika-labs

Use when the user asks for podcast or a task matching the examples below. Two-host podcast video for any URL or free-form topic — 1 minute, 4 acts × ~15s, native multi-shot dialogue, optional voice cloning for Host A. Use when the user asks to "make a podcast", "podcast about [thing]", "podcast review of [url]", "two-host explainer", "interview-style clip", "two people talking on camera", "I/me and X talk about Y", or "interview with [persona] about [topic]". Native audio is the deliverable; captions are skipped by default because podcast dialogue mistranscribes domain terms.

待分类402个月前更新

Persona Builder

pika-labs

Use when the user asks for persona builder or a task matching the examples below. Whip a person into a more marketable shape online. Read their socials and the way they talk, then deliver real talk about where the money is, what's holding them back, and what to fix — then ship the designed multi-page Influencer Persona PDF + a persona.md folder kit that downstream skills (ugc-ads, podcast, founder-product-video, app-sizzle, app-store-screens) can consume. Input is the person themselves: socials, camera roll, taste URLs, a selfie video, or start-from-scratch answers. Visual/PDF stages still require real curated imagery: user-provided photos, or clean public-feed frames from a supported social handle. Output is a self-contained kit + a roadmap to actually become that persona online. Trigger phrases: "build my influencer identity", "make me a creator brand", "personal brand for [niche]", "build my online persona", "influencer persona.md", "I want to be an influencer", "make my creator identity", "persona.md for

待分类402个月前更新