Generate AI Marketing Videos From Claude Code: The Arcads Skill Pack
A Claude Code skill pack that connects to the Arcads creative stack — Seedance 2.0, Sora 2, Veo 3.1, Kling 3.0, and more — plus a 37-template static image-ad library and prompting guides for each model.
Last updated: June 15, 2026
A Claude Code skill pack that connects to the Arcads creative stack — Seedance 2.0, Sora 2, Veo 3.1, Kling 3.0, and more — plus a 37-template static image-ad library and prompting guides for each model.
This guide is reviewed for clarity, service accuracy, and AI-search readability. The next quarterly content review is tracked internally before unsupported metrics or client proof are added.
What this skill pack does
The Arcads skill pack connects Claude Code to the Arcads AI creative platform — a stack of video and image generation models purpose-built for marketing ads. From inside your Claude Code session, you can generate UGC-style product videos, premium product reveals, studio lookbooks, feature walkthrough demos, and static Meta image ads without leaving the agent.
The agent handles API calls, polling, prompt engineering, file organization, and cost confirmation.
Getting started
bash# Clone the repo git clone https://github.com/krusemediallc/arcads-claude-code.git cd arcads-claude-code # Run setup (installs deps, saves your Arcads API key, verifies connection) ./scripts/setup.sh
Open the folder in Claude Code. A
SessionStart.envMASTER_CONTEXT.mdThe model stack
- Seedance 2.0 — flagship video model. 4–15s clips, native audio, image-to-video or video-to-video, reference images, multiple shot styles.
- Sora 2 — longer-form cinematic content via OpenAI's model
- Veo 3.1 — Google's model, strong on photorealistic motion and product shots
- Kling 3.0 — smooth motion, good for lifestyle and fashion content
- Grok Video — variety; use when other models don't fit the aesthetic
- Nano Banana 2 / Pro / Edit — short-form, fast-turnaround and edit-from-existing-asset workflows
- ChatGPT Image 2 — static image ads, stdlib-only
- OmniHuman / Audio-driven — avatar and talking-head formats for spokesperson-style ads
Five Seedance 2.0 prompt formulas
The skill ships with five prompt formulas tuned for specific ad types:
- UGC selfie-style product review — prompt
"Make a 12-second Seedance UGC video — woman in a kitchen, holding the product, says she stopped buying [competitor]" - Premium product reveal (no person) — prompt
"Premium reveal of [product] — dark void, text narrative, hero rotation" - Product hero with elemental effects — prompt
"Seedance product hero — water splash, mist, slow rotation" - Studio lookbook with voiceover — prompt
"Studio lookbook of [product] — multi-look, polished, with voiceover script" - Feature walkthrough demo — prompt
"Seedance feature walkthrough — fast-paced, show off [features]"
Static image ads
Beyond video, the skill pack includes a 37-template static Meta image-ad library plus Pixar-style and claymation animated ad pipelines. The image-ad generators (
chatgpt-image-adnano-banana-image-adimage-ad-clonePrerequisites by workflow
- Python 3.10+ — required for everything
- ffmpeg — Pixar-style ad, claymation, caption-video: prompt
brew install ffmpeg - jq — several bash scripts: prompt
brew install jq - Node.js — caption burn-in: prompt
brew install node - whisper — caption transcription: prompt
pip install openai-whisper
Frequently asked questions
Do I need an Arcads account? Yes. The skill calls the Arcads external API.
./scripts/setup.sh.envHow does cost confirmation work? The agent prompts for confirmation before any generation workflow. Video generation cost varies by model, duration, and resolution — the skill shows estimated cost before executing.
Can I use this in Cursor as well as Claude Code? Yes. The same skills are exposed at
.cursor/skills/