CommunityVideo y animacióngithub.com

LeechQAQ/poster-to-video

An open Agent Skill that turns one poster into a 15-second vertical promotional video workflow and Seedance-style prompt.

Compatible conClaude Code~Codex CLI~Cursor
npx skills add LeechQAQ/poster-to-video

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentación

Poster to Video

Turn a single poster or promotional image into a 15-second 9:16 marketing video package.

Package Layout

poster-to-video/
  SKILL.md
  forms.md
  reference.md
  scripts/
  resources/

Read only what is needed:

  • Use forms.md for the generation package, submit checklist, and final report format.
  • Use reference.md for detailed creative rules, visual-analysis prompts, Seedance 2.0 conventions, and quality gates.
  • Use resources/prompt-template.txt when writing a copyable prompt file.
  • Use scripts/check_seedance_prompt_lengths.ps1 before submitting a Seedance-style prompt.

Defaults

  • Output target: 15 seconds, 9:16, 720p or higher, audio on.
  • One run uses one main reference image unless the user provides more assets.
  • Final video-model prompt should be Chinese when targeting Seedance 2.0.
  • Do not invent prices, discounts, certifications, brand endorsements, conversion data, platform UI, hardware specs, or QR codes.
  • If the input contains a QR code, preserve only the original QR code as part of the source image or a protected crop; do not redraw or replace it.
  • Do not add new on-screen subtitles, scrolling subtitles, bottom caption bars, random watermarks, or extra logos unless the user explicitly asks.
  • Prompt length should stay under 3800 characters for Seedance-style submission.
  • Preserve important visible text with crops, cards, masks, or highlights instead of asking the model to redraw dense Chinese text.
  • Always include a no-garbled-Chinese constraint for generated or preserved Chinese characters.

Workflow

  1. Inspect the input image before writing any video prompt.
  2. Classify the source: campaign poster, advertising screen, template page, product poster, digital signage scene, kiosk display, sale banner, or other promotional asset.
  3. Extract visible facts: title, subtitle, small text, numbers, CTA, buttons, icons, cards, products, QR code status, scene, layout hierarchy, color palette, and style.
  4. Build an information-priority table:
    • Must preserve.
    • Can weaken or omit.
    • Should not read aloud.
    • Must not invent.
  5. Choose one video direction: campaign launch, template reuse, store attraction, product scene, seasonal sale, feature demo, or quick conversion CTA.
  6. Define one core benefit and one visual anchor from the source image.
  7. Write a 5-7 shot progression for 15 seconds. Each shot must have an integer-second range and one business-purpose motion.
  8. Write the generation package using forms.md.
  9. Write the final Seedance-style Chinese prompt using reference.md or resources/prompt-template.txt.
  10. Run the strict prompt preflight when a prompt file exists.
  11. If the user provided a configured video-generation runner, submit the source image and prompt. Otherwise, return the ready-to-submit prompt and provider-neutral instructions.

Hard Preflight Checks

Keep these checks even when the user says to run quickly:

  • Prompt is under 3800 characters unless the target provider allows more.
  • Prompt uses Chinese field labels for Seedance-style submission.
  • No banned English labels such as Section, Scene Content, Dialogue, Subtitle Content, Voiceover Reading, SFX, Camera/Transition, or Negative Constraints.
  • No standalone 参考图解析 section in the final prompt; source analysis belongs in the relevant @图片N = ... mapping line.
  • Every shot heading includes an integer-second range, such as 分镜一(0-2秒).
  • Any voiceover line uses {} and includes toned pinyin for every pronounced syllable if the target TTS requires it.
  • No new subtitle fields unless the user requested visible subtitles.
  • No raw URL, local path, or asset ID appears inside the prompt prose.
  • Every shot that uses a source asset references the matching @图片N.
  • Prompt includes a strong rule against garbled Chinese, mojibake, pseudo-Chinese, wrong characters, random strokes, and deformed text.

Commands

Validate a Seedance-style prompt:

& '.\skills\poster-to-video\scripts\check_seedance_prompt_lengths.ps1' `
  -PromptPath '.\outputs\poster-video-prompt.txt' `
  -StrictChineseSeedance

Install this skill locally from the repository root:

.\install.ps1

Final Response

Use the final report format in forms.md.

If video generation was actually submitted and a playable video URL was returned, include the raw URL as visible text. Do not hide it only behind a Markdown link.

Skills relacionados

higgsfield-ai/higgsfield-soul-id

Train a Soul Character — a personalized model on a person's face that Higgsfield uses for identity-faithful image and video generation. Use when: "create my Soul", "train my face", "make my digital twin", "build me an avatar", "learn my appearance", "create a character of me", "set up identity for video", "I want my face in generated images". Chain: train Soul (one-time, returns reference_id) → use in higgsfield-generate via `--soul-id <id>` with models like `text2image_soul_v2` or `soul_cinema_studio`. NOT for: one-shot face swaps (use higgsfield-generate with --image), named-character / non-photo avatars (use higgsfield-generate with prompt).

community

huytjuh/AFKOps

The ultimate AI Agent Workflow for gaming content creators. Deploy mini PC grids to generate 24/7 streams, VODs, and ad revenue across YouTube, Twitch, and more—while you're away from keyboard.

community

wohuishuo/cc-video-pipeline

Claude Code skills 视频创作流水线

community

gabros20/youtube-relay-mcp

Code-execution YouTube tool for AI agents — CLI (ytrelay) + MCP shim + Claude Code skill. Search, transcript, metadata, embeddable ID. Powered by youtubei.js.

community

Apizziii/OpenClaw_RH_Skills

Provide multimedia generation and AI app integration for OpenClaw using RunningHub’s extensive API endpoints in image, video, audio, 3D, and text.

community

heygen-com/remotion-to-hyperframes

Translate an existing Remotion (React-based) video composition into a HyperFrames HTML composition. Use ONLY when the user explicitly asks to port, convert, migrate, translate, or rewrite a Remotion composition as HyperFrames (e.g. "port my Remotion project to HyperFrames"). Do NOT use when (a) authoring a NEW HyperFrames composition (even if A/B-testing a Remotion video); (b) Remotion is mentioned in passing; (c) Remotion code is shared as reference, not for translation; (d) the user wants "the same video as my Remotion one" without explicitly asking to migrate the source — treat as a fresh HyperFrames build. When in doubt, default to the `hyperframes` skill. Detects unsupported patterns (useState, useEffect side effects, async calculateMetadata, third-party React component libraries, `@remotion/lambda`) and recommends the runtime interop escape hatch instead of a lossy translation.

community