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 (

/pika:ugc ads Parameters Param Default Notes url required product URL — drives category detection and beat substitution avatar url built in fallback persona portrait URL; fed as @Image1 reference. When omitted, the skill uses a pre generated Pixar style female creator portrait provider seedance seedance: strong at UGC selfie / talking head POV with native lip sync, multi segment in single prompt, supports 3:4. kling: explicit shots[] , 9:16/16:9 only aspect ratio 9:16 3:4 is seedance only (kling rejects 3:4) variants unset Optional comma list for shared generation exports. Supported: 9:16 , 16:9 , 1:1 . Keeps the expensive UGC render shared, then reframes the final stage. category auto HAUL / APP / FOOD / BEAUTY / FITNESS / TECH ; auto picked from URL captions true TikTok style word chunked captions burned on top of the final video Cost transparency gate Before any paid MCP call, call identity balance({verbose: true}) once. Surface the current balance, recent burn rate, and remaining runway, then gate the run with this exact message: Estimated cost: about 4,000 credits (~$40) for a typical Seedance UGC ad with fallback/retry budget. This exceeds $5, so Reply proceed to continue or cancel to stop. Do not call any paid MCP tool until the user replies proceed . If the user replies cancel , stop without generating. The gate runs after the product URL is known and before avatar analysis, screenshot capture, video generation, captions, or paid retries. Runtime expectations Typical end to end run: 6–12 minutes . Breakdown: Step 1 (WebFetch) + Step 3 (capture website screenshot): ~10–30s Step 7 ( generate reference video ): ~3–5 min for seedance, ~5–7 min for kling Step 7b/c (cartoonize + retry): adds ~1–2 min if seedance moderation rejects the avatar Step 8 (deterministic brand/spec overlays): 1 2 edit text overlay calls, ~30s–5 min total Step 9 (captions): single add captions call, ~30s–5 min (transcribe + burn in one shot) If the run exceeds 15 min without progress, something is wrong — inspect the tool reported generation status and error message. Pre generation wall clock guard Start a timer at skill start once the product URL is available and the cost gate has passed. Time spent waiting for the user's proceed reply is not prep time and must not trigger this guard. The first paid generation call is generate reference video , the long pole paid stage, and it must be invoked within 5 minutes of skill start. If you have not invoked generate reference video within 5 minutes of skill start, stop before any paid generation call and report failed pre generation timeout with what you have so far: fetched product facts, chosen category, avatar source, screenshot status, draft dialogue, and the exact blocker. Do not keep refining script wording, prompt grounding, or shot order. Print a single line progress checkpoint after each prep stage and right before the paid generation call: Stage 1/3 done — product fetched and categorized. Stage 2/3 done — avatar and screenshot ready, composing dialogue. Stage 3/3 done — prompt locked, calling Seedance now. Script and prompt iteration is maximum 2 passes. After the max 2 passes, ship what you have to generate reference video ; do not continue polishing the hook, punchline, or screen close up wording. Long running task status polling When any long running generation or edit call returns a task id with or without an initial status, including {task id} , {task id, status: "queued"} , or an initial queued , running , or processing status, record the task id and start time immediately. Call task status({task id}) in a tight loop until terminal ( completed failed cancelled ). No manual sleep and no Bash polling; the worker holds each status call open. Emit ONE visible progress line every 60s while status is queued , running , or processing : Seedance i2v queued for {N}m {S}s... still processing . Replace the provider/stage label when polling Kling, GPT image 2, caption, or edit tasks. On completed , unwrap the returned result URL and continue. On failed or cancelled , surface failure to the user with task id , status, and the last status message. After 15 min total from the original submit, call task cancel({task id}) if the task is still non terminal, then surface failure to the user. If cancel reports the task is already terminal, call status once more and report that terminal result. Do not submit a duplicate request while the original task is still queued , running , or processing . Async polling budget: one active task with one 15 min polling window. After the polling cap is exhausted, cancel the task and surface failure instead of submitting another paid render unless a later step explicitly allows a separate capped retry after the original task is terminal. Engine choice: Seedance default, Kling fallback Default to Seedance for UGC selfie/talking head ads because it handles native lip sync, single prompt multi beat pacing, and optional 3:4 output well. Use Kling when the caller explicitly passes provider=kling , or after Seedance exhausts the capped cartoonized retry only if the user chooses Kling from the stop message. Kling's tradeoff is stricter aspect ratio support but a separate moderation path and explicit shot segmentation. Steps 0. Resolve input (empty args menu) Strip flags and key=value parameters from $ARGUMENTS . If no product URL remains and there is no usable product URL in prior context, print this menu and stop: Which product should the UGC ad promote? Required: Product URL — page to fetch for product name, category, visual references, and language Optional: avatar url= , provider=seedance kling , aspect ratio=9:16 3:4 , variants=9:16,16:9,1:1 , category=auto HAUL APP FOOD BEAUTY FITNESS TECH , captions=true false . If the product URL is present, skip this step silently. 1. Fetch + categorize WebFetch the URL: pull product name , brand name , value prop, brand color, product form, packaging, hero copy, target user, category, and the primary language of the page . Use category= if passed; else trust the WebFetch signal; fall back to HAUL for physical, APP for digital. Build two grounded fact lists from the fetched page: Every numeric spec claim ( W , mAh , % , minutes, ports, price, dimensions, counts, charging speeds, battery size, rankings) must come from visible source page text and include a source quote . If the source page does not visibly support a number, leave it out of grounded specs and claims allowlist . Do not infer specs from product category, model name, common knowledge, or competitor pages. 2. Resolve the avatar (fallback to built in if missing) If avatar url was passed → use it as is. If NOT passed → use this built in fallback: Pre generated 3D animated Pixar style portrait of a young female creator — pre cartoonized so seedance moderation accepts it directly, neutral enough to fit any category. Note in the final summary that the fallback was used so the caller knows to supply their own portrait for persona consistency next time. 2.5 Avatar type probe for creator portraits Run this probe before Step 7 and before any paid generate reference video call. It applies to caller supplied avatar url , the built in fallback, or any creator portrait chosen as @Image1 . The built in fallback is already a non IP stylized creator, but still document its source in the final summary. Call analyze media once: Route from the result: recognized IP / copyright risk STOP only when avatar type is "recognized ip" , or recognized character names a specific character (for example "Batman" ), or when both moderation risk is "high" and recommendation is "reject" . Treat recognized character: null , empty string, "none" , "unknown" , "n/a" , and low/medium moderation risk as not enough to stop by themselves. Run this check before the real/stylized routes. A chibi Batman is still Batman even when avatar type is stylized / illustrated. real human / AI generated realistic proceed normally. stylized / illustrated proceed with a visible warning that stylized avatars may be accepted by Kling but can be inconsistent under Seedance moderation; continue only if the user supplied or accepted that avatar. trademarked / copyrighted STOP before generation. Surface this message: The avatar appears to be a trademarked character ([X]). Most video providers will moderate this and refuse to generate. Pass avatar url=<real looking photo url to override. 3. Capture the product screenshot (best effort) Call capture website with mode: "screenshot" . Use mobile=true for handheld product categories (APP / FITNESS / BEAUTY) so the captured page renders as a portrait phone screen; mobile=false for desktop context categories (HAUL / TECH / FOOD). If the call fails (timeout, browser pool down), retry once . If still failing, proceed without the screenshot — the skill is degraded but functional. The close up beat then describes the page from prose only and Beat 2's reference images is just [avatar url] . Capture URL → screenshot url (or null). 4. Compose the prompt The full prompt is a single multi beat string passed to one generate reference video call. Structural prose (not markdown bullets). Every beat has a Says: "..." line for lip sync. Pacing target ~5.5–6 words per second across the whole 15 second ad (≈85–90 words total). @Image1 is the avatar, @Image2 is the screenshot when available. Write all Says: "..." lines in the language detected from step 1's WebFetch. Both seedance and kling lip sync handle multilingual; if the product page is Chinese / Japanese / Spanish / etc., the dialogue should be in that language. Hook archetypes from step 5 are language agnostic — adapt the rhetorical move to the language's natural register. Spec grounding rule: forbid inventing numbers. Any spoken or visual number/unit claim must appear verbatim in claims allowlist . If a number is not in claims allowlist , rewrite the line qualitatively ("charges fast", "multiple ports", "big battery") or omit the claim. Do not say "50% in 28 minutes", "3 ports", "140W", prices, counts, or time windows unless that exact claim is source backed. Brand/spec text rendering rule: Do not ask Seedance or the video model to render the brand wordmark, product wordmark, packaging label, or spec text from prose. Video model text comes out garbled. The prompt may show @Image2 as a reference, but any new brand name or spec copy that must be readable is added later by the deterministic Step 8 overlay. Screen close up beat — exactly one across the ad, position is dialogue driven: Place the screen close up on whichever JUMP CUT (1, 2, or 3) the reveal line lands on. Most ads put it on JUMP CUT 2; if the narrative needs it earlier or later, JUMP CUT 1 or JUMP CUT 3 is fine. Pick by content, not by slot number. The screen close up beat shows @Image2 exactly as is and includes ONE finger point gesture (a single finger entering from the frame edge, pointing at the hero text or product — no tap, no swipe, no scroll, no hover on CTA). The point gesture is the only screen interaction in the entire ad. The other JUMP CUTs are wide POV reaction beats: hands stay on knees, on the bed, or at sides. Trust @Image2 — when the product page is shown, reference the image; do NOT describe its UI in prose. Describing UI triggers the model to invent extra panels / dropdowns / sidebars / animations. Reference the image; trust it. 5. Category essences Each essence is the brief you read before composing the 5 beats. Pick one from category in step 1 and write the actual Says: "..." lines tailor

ugc-ads