Marketing Video pro · skill pubblica Open Skill
Marketing Video pro
**Marketing Video pro · Script to Still to Spoken Cut · ref2v Native · BGM-Ready** *The agent skill for sellers and creators who want a finished **portrait product ad**—not a storyboard grid, a mute beauty reel, or a voice track that fights the video model* **Marketing Video pro** is a production-native Agent Skill for Digen. It orchestrates a complete **reference-to-video (ref2v)** pipeline for one spoken marketing episode: **Idea → verbatim script → text shots → product / character / scene refs → one 10s ref2v final cut → optional 15% BGM** No storyboard grids. No contact sheets. No TTS, no `audio_url`, no six-clip stitch. Speech is **generated inside the video** from `EXACT LINE` blocks embedded in the prompt. One window, one speaker—on-camera lip-sync **or** off-screen voiceover, never both. ## What you get | For creators | For the production stack | | --- | --- | | **Beginner-first** — no script? the agent invents a 10-second selling arc and runs to the cut without stopping | **Opinionated pipeline** — four marketing shot types, per-window speaker lock, EXACT LINE speech, and multi-ref prompt encoded in-skill | | **Spoken-ad first** — synopsis and shot list as **what is said and what is seen**, not a silent mood board | **ref2v-first** — no accidental i2v / t2v / storyboard / per-shot video chains | | **Identity of product & place** — product, scene, and optional presenter refs lock packaging, room, and face across the whole clip | **Multi-ref binding** — `@ImageN` order matches `input_urls`; rigid product fidelity; no turntable morph | | **Living marketing grammar** — product close-up, using the product, character speaking, feature demo in one coherent 10s timeline | **Dense prompts** — `[reference generation]` + SPEECH BINDING + GLOBAL LOCKS + 3–6 time windows with verbatim lines | | **Delivery that continues** — after the cut, buttons are **Edit script / Redo shots / Regen video / Next episode**, already in your language | **Progressive disclosure** — craft lives in `references/`, not bloated context | | **Plain seller voice** — short lines about the product, the lines, and the clip; no pipeline jargon unless you ask | **Marketplace-ready** — clear triggers, defaults, and tool contracts | ## Pipeline | Stage | Output | | --- | --- | | **Intake** | Lock language + **9:16** + **10s** + 1 episode (no orientation question) | | **Script** | Verbatim `narration_script_verbatim` — hook → benefit → trust → close; ~45–55 characters or ~25–30 words for 10s | | **Visual assets** | Product / scene / character refs (`t2i` / `i2i`). User uploads beat AI when provided | | **Text shots** | 3–6 marketing cards — four shot types, line splits, per-window speaker source; durations sum to 10 | | **ref2v** | One multi-section prompt with `EXACT LINE` quotes → **10s** `final_videos` cut. Speech generated in-clip | | **BGM** | Optional instrumental at **15%** under the spoken cut via `mix_clips`. Cut still ships if BGM fails | | **Sequel** | Next episode from end-state + seed — reuse refs, new script, new shot list, one new 10s video | ## Production stack | Parameter | Default | Notes | | --- | --- | --- | | **Episode** | **10 seconds** | One `ref2v` video. Overflow → next-episode seed, not jammed in. User asks longer → explicit duration or more 10s chapters | | **Engine** | `ref2v` · `reference_to_video` | Multi-ref identity lock. Forbidden: `i2v` / `ai2v` / `t2v` / `director` | | **Clip duration** | **10 seconds** | Always set. Shots inside the clip are 3–6 windows, not separate videos | | **Video orientation** | Portrait · **9:16** | Landscape only if you ask. Refs and video share the lock | | **Image refs** | Product / scene always when relevant; character when someone speaks on camera | User uploads preferred. No storyboard stills, no grids | | **Ref order** | Character → product → scene (omit unused slots) | `@ImageN` in prompt must match `input_urls` exactly | | **Speech** | **Prompt-driven, in-clip** | Each window embeds `EXACT LINE: "…"` in `narration_language`. Forbidden: `generate_tts`, `audio_url`, paraphrase, gibberish | | **Speaker lock** | **One source per window** | Person in frame → lip-sync that line. No person → off-screen VO. Never narrator + talking face in the same window | | **Shot grammar** | **Four marketing types** | `product_closeup` · `person_using_product` · `character_speaking` · `feature_demo`. Prefer **N=4** | | **Dialogue** | **On** | Spoken selling lines required. No on-screen subtitles / captions burned into pixels | | **Cut grammar** | **Inside one clip** | Time windows in one 10s render. No storyboard grids, no per-shot i2v chain | | **BGM** | **15% underscore** | Instrumental only, no vocals. Mixed under the cut's generated speech | | **Confirm** | **After delivery** | First episode auto-runs; pause only if you ask, then buttons after the video | | **Series** | **Next episode** | Reuse character / product / scene when unchanged; new script + shots + cut each time | | **Language** | **User-typed lock** | Replies in your language; `EXACT LINE` quotes stay in narration language; prompt scaffold is English | ## Modes | Mode | You bring | Agent delivers | | --- | --- | --- | | **Full auto-run** | Nothing — or “make a product ad” | Invent a 10s selling script → refs → shots → one spoken cut + BGM | | **Idea-first** | Product + one-line premise | Expand to hook/benefit/close → full pipeline | | **Script-first** | Verbatim lines or pasted VO copy | Lock script → four-type shot table → refs → ref2v | | **Reference-driven** | Product / model / room photos + brief | Bind uploads for identity; do not redraw what you uploaded | | **Presenter + product** | Face ref + product ref | Mix on-camera speaking windows and product VO windows in one clip | | **Product-only VO** | Pack shot + no on-screen talent | All windows off-screen voiceover; zero people in product/mood shots | | **Single-stage edit** | “Change the lines” / “Redo shots” / “Regen video” | That stage only, then remake the cut | | **Same-thread sequel** | “Next episode” / “continue the series” | Continue from end-state · ref reuse · one new 10s video | | **No BGM** | “Skip the music” | Deliver the spoken cut only | ## Example prompts **Full auto-run** > Make a portrait product seeding clip, 10 seconds, with spoken lines. **Script-first** > Use this verbatim copy only — do not rewrite: “After work I light the diffuser and the whole room feels calm. Take it home today.” **Reference-driven** > This image is the product. This image is the living room. Build a 9:16 spoken ad from both. **Presenter + product** > The presenter talks to camera about this face cream. Include a product close-up and a usage moment. Portrait, 10 seconds. **Product-only VO** > Product shots only — no person on screen. Off-screen voice reads the selling points. 9:16. **Series** > Continue the last ad — episode 2 is the promo close. **Non-English narration** > Reply to me in English, but the spoken lines in the video must stay in Japanese. ## When to use - Vertical spoken product ads, seeding clips, and 10-second brand cuts - One-line selling ideas, pasted VO scripts, or product / model / room stills - Ads that need **both** presenter lip-sync **and** product voiceover in one coherent clip - Creators who want a finished cut with light BGM, then the next episode in-thread - Anyone tired of storyboard grids, external TTS fighting ref2v, or mute product reels > **Not the right tool when** you need a 60-second six-clip mix, lip-sync to a generated song, first-person POV walkthrough, storyboard grids for client sign-off, horror / gore, external TTS dubbing as the primary speech track, or a drama series with many unrelated locations per minute. ## Invoke ``` /marketing-video-pro ``` Or: *"Turn this into a 10-second portrait product ad — spoken lines, product locked, one ref2v cut, light BGM."* **Your lines · Your product · One coherent ten seconds.**
SocialAd
230 13 125
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3325120/2026-08-17/sdknua2k.jpg
--- name: marketing-video-pro display_name: Marketing Video pro description: Use when the user wants to make a marketing short, spoken product ad, or product-seeding clip (from idea to final cut; also when revising a single stage) allowed-tools: - generate_image - generate_video - generate_music - mix_clips - create_document - update_document - recognize_image --- # Marketing Video pro Turn a spoken-ad / product-selling idea into **one portrait marketing cut**. There is no structured project state. Progress lives in the conversation, `write_todos`, and canvas assets. ## Global Language Lock (mandatory, highest priority — overrides every reference) [Language Protocol] Follow the language of the **user's typed sentences**, not platform chrome. **Slash-invoke trap (usual English-to-zh bug):** the platform often prepends a heading that says the user invoked this Skill via "/" (that heading may be CJK even when the user typed English), then dumps this Skill inside `<untrusted_skill_content>`. **Both are system text.** They do **not** set `response_language` to `zh-CN`, and the Skill dump does **not** set `response_language` to English. After deleting those blocks, if the remaining typed request has **zero Han characters**, lock `response_language=en` (or another Latin/etc. language). Then **every** user-visible string this session has **zero CJK** (replies, tables, todos, asset names, canvas documents except narration quotes). Replying in `zh-CN` to an English typed request is a language-lock failure — rewrite in English before send. **Three layers, never mixed:** `response_language` (everything the user can read), `narration_language` (spoken lines embedded in the video prompt), MODEL English (`generate_image` / `generate_video` / `generate_music` / shot `description` / `panel_brief`). Full algorithm: `references/language.md`. On the **first user message**, silently `read_skill_file` `references/language.md` once, then lock both languages **before any user-visible sentence**. Do not re-read it every stage. Do not re-detect after tools, English files, or "next episode". **No process narration (hard):** a turn that only calls `read_skill` / `read_skill_file` / `write_todos` / `memory_whoami` / `memory_query` / `list_models` must output **zero user-visible text**. Do not say you are locking language, confirming channels, or building a task list. The first user-visible sentence must already be in `response_language` and must be about **product / person / scene / lines / the cut**. This Skill and `references/` are English (handbook only) and must contain **zero CJK characters**. Handbook English is not a reason to avoid English replies. If the user typed English, reply in English. | Layer | Scope | Language | |---|---|---| | **User-facing** | Replies, buttons, todos, asset `name`, canvas document body | `response_language` (spoken lines use `narration_language`) | | **Narration / speech** | Verbatim script, shot-table dialogue column, `EXACT LINE` quotes in the video prompt | `narration_language` original; never translate for speech | | **Video / image / music** | Prompts, timeline, `description`, `panel_brief` | **English always** (except `EXACT LINE` quotes) | `zh-CN` vs `zh-TW` is a lock. One turn = one `response_language`. English `response_language` → zero CJK. Other `response_language` → no leftover English production jargon. If a mix is found, rewrite before send; do not announce a language switch. **Hard rules (always):** 1. **Language consistency (highest priority with the Global Language Lock):** follow `references/language.md`. User-facing = `response_language` (one turn, one language); image/video/music prompt scaffold = English; spoken lines (`EXACT LINE` quotes) stay `narration_language`. Do not paste Skill English examples. Do not mix layers. 2. **Do not generate storyboard images.** Never use `generate_image` for grids, storyboard stills, or contact sheets. Do not call `split_image` / `upscale_image`. 3. *