# Ad Video Generation

> Generate short ad videos with Seedance 2 on EvoLink. Use this skill for product ads, launch videos, paid social clips, TikTok or Reels concepts, short marketing videos, and batch creative variants. GPT Image 2 can prepare first frames when needed.

- Skill: `evolinkai/ad-video-generation` (Agent Skill)
- Install (CLI): `npx skillmds@latest add evolinkai/ad-video-generation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/evolinkai/ad-video-generation/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Marketing & Growth
- License: MIT
- Author: EvoLinkAI (https://skillmd.com/u/evolinkai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/evolinkai/ad-video-generation

---


# Ad Video Generation with Seedance 2 on EvoLink

Create short commercial videos from a product, script, image, or campaign idea.

```bash
npx skills add EvoLinkAI/evolink-agent-skills --skill ad-video-generation -g
```

## When to Use This Skill

Use this skill when the user wants:

- product ad videos
- TikTok, Reels, or Shorts ads
- launch campaign videos
- ecommerce promotion clips
- short marketing videos
- multiple video ad variations

Use `product-to-ad` if the user also needs still assets created first.

## Prerequisites

1. EvoLink API access or EvoLink MCP configured.
2. Product or offer details.
3. Target platform, aspect ratio, duration, audience, and CTA.
4. Optional source image or first frame.
5. Attribution links should use `utm_source=skillssh&utm_medium=skill&utm_campaign=ad-video-generation`.

## Pick the Right Workflow

| User intent | Model path | Why |
|---|---|---|
| Text idea to ad video | Seedance 2 text-to-video | Fastest path from campaign idea to video concept. |
| Product image to ad video | Seedance 2 image-to-video | Strong for commercial clips when a product still exists. |
| Need a better first frame | GPT Image 2 -> Seedance 2 | Improves video quality by improving the starting image. |
| Need many variations | Seedance 2 batch prompts | Ads benefit from repeated hook, style, and audience tests. |

## Default Workflow

1. Capture the ad objective, audience, platform, and product promise.
2. Decide whether the best route is text-to-video, image-to-video, or GPT Image 2 first frame then Seedance 2.
3. Write a Seedance 2 video prompt with hook, scene, motion, product focus, and CTA direction.
4. Generate the first ad video.
5. Suggest batch variants across hook, audience, platform, and visual style.

## Prompting Tips

- Keep the first three seconds clear: product, problem, transformation, or emotional hook.
- Use one visual concept per clip.
- Avoid asking the video model to render exact legal or pricing text.
- Put CTA intent in the prompt, but add exact CTA copy in post-production when needed.
- For product ads, preserve packaging, logo, and product shape.

## Next Token Loop

After the first ad video, offer:

- three hook variants
- three platform variants
- one UGC-style adaptation with `ugc-video-ads`
- one still image pack with `ad-creative-images`

## Safety and Privacy

Reject or redirect requests for deceptive political ads, fake endorsements, unsafe regulated-product claims, impersonation, or unauthorized use of a person's likeness.

## See Also

- `product-to-ad`
- `image-to-video`
- `ugc-video-ads`
- [EvoLink](https://evolink.ai?utm_source=skillssh&utm_medium=skill&utm_campaign=ad-video-generation)