UGC Video Ad Skill for AI Agents

The full ad

One prompt in. A finished ad out.

Nothing below was edited by hand. One sentence for a real brand, and the skill wrote the script, cast the creator and rendered each shot with its own spoken audio.

Make a UGC ad for Som Sleep. Problem-solution angle, you cast the creator.

Start creating Install the skill

Your product, your angle, the same pipeline.

Just ask

Say the word, it runs.

The skill fires on natural requests like these. No commands to memorize.

“Make a UGC ad for my product”“Turn this product into a TikTok/Reels ad”“Create a creator-style testimonial video”“Product-in-hand demo video”

The pipeline

What the skill does for you.

1
Pick brand and angle
The skill reads your Advibly brand, product, and research brief, then picks one of five preset angles: testimonial, car / on-the-go, unboxing, lifestyle demo, or problem-solution.
2
Cast an AI creator
It generates a realistic, unretouched creator reference with gpt-image-2, using a candid-iPhone-photo prompt formula tuned for realism, not model gloss.
3
Write a direct-response script
A 5-shot shot list with dialogue, paced at roughly three words per second, hook first, product name held back until shot three, punchy CTA. Script edits are free, so you lock it before spending on video.
4
Build the storyboard
One start frame per shot with the creator and product as references, presented as an approval gate. A frame tweak is one cheap image; a bad clip is an expensive render.
5
Animate every shot
Each approved frame becomes a clip with Gemini Omni Flash image-to-video, native audio and all, so the creator speaks the line in the clip itself.
6
Stitch the finished ad
With a shell, it downloads the clips and concats them in order into one vertical ad. Without one, it hands back the clips in order with CapCut stitching guidance, then can publish to a connected account.

You direct. The agent shoots.

Approve the lines and the storyboard, and every clip after that renders on its own.

Start creating

Casting

A new creator for every product.

The skill does not reuse one stock face. It matches a creator to what you sell, then prompts for a candid iPhone photo with flyaways, real skin and imperfect light. That creator is a reference on every storyboard frame, so one person carries the whole ad.

Eight creator archetypes matched to your product category, or describe your own
Phone-photo realism on purpose, never model gloss
The same face and outfit from the hook to the CTA

Start creating Install the skill

Coffeeverse

Arata

AG1

The model stack

The right model for every shot.

GPT Image 2 is the default for the creator and the five storyboard frames, because it keeps your label readable and holds one face across references. Gemini Omni Flash is the default for the clips: it animates each approved frame and generates the audio in the same pass, so the line is spoken inside the clip. Every video model here generates the speech in the clip, and your agent runs the newer ones, like GPT Image 2.5 Sunburst, Gemini Omni Flash 1.1 and Seedance 2.5, when you ask.

What you get

A finished vertical (9:16) UGC video ad
Five talking clips, one per shot, with spoken dialogue
A direct-response script and approved storyboard
Everything filed in your Advibly library, ready to publish

Start creating Read the source

GPT Image 2.5 SunburstNew

Best-quality creator portrait and storyboard frames when detail matters

GPT Image 2.5 FlareNew

The fastest storyboard drafts while you settle the look

GPT Image 2

Default: casts the creator and draws the five storyboard frames

Nano Banana 2

Storyboard frames up to 4K for a sharper start frame

Gemini Omni Flash 1.1New

Talking clips that take reference clips to keep the creator's face and voice across shots

Seedance 2.5New

One longer take, up to 30 seconds, with the line spoken in the clip

Wan 3.0New

The lowest cost per second with native audio, for testing many hooks

Gemini Omni Flash

Default: turns each approved frame into a talking clip with native audio

Kling 3

Talking clips with native audio, including square 1:1 cuts

Install

Add it in seconds.

Install the skill, then make sure the Advibly MCP is connected so it has tools to call. Works with every skills-capable agent.

Claude Code

Claude

Cursor

Codex

VS Code

OpenClaw

Hermes

ChatGPT

Connect the Advibly MCP first

One click, no terminal

Download .skill Download .zip

Drop the .skill file into Claude → Settings → Skills, or unzip the folder into ~/.claude/skills.

One command (any agent)

npx skills add advibly/skills -s ugc-ads

Works with Claude Code, Claude Desktop, Cursor, and any agent that supports skills. Pulls straight from GitHub.

Manual (Claude Code)

git clone https://github.com/advibly/skills.git
cp -R skills/ugc-ads ~/.claude/skills/

Or in Claude Desktop: Settings → Skills → Install and select the skill folder.

Bundled references

The playbook.

Each skill ships with reference docs the agent loads on demand, so the craft is baked in.

prompting.md

Style modifiers, start-frame rules, dialogue pacing, and hallucination workarounds.

angles.md

Shot-by-shot breakdowns for all five preset ad angles.

characters.md

Creator prompt templates for eight archetypes, tuned for gpt-image-2.

Honest limits

What to watch for.

Voice drifts slightly between clips (each render synthesizes its own). For one continuous voice, use the talking-actor tool instead.
Product label text can garble when the product is small or the scene is busy. Keep the product large and close in product-forward shots.
Eating and drinking shots hallucinate. Use hold-and-show endings for consumables.
One simple action per shot. Complex hand-object interaction warps.

Questions

Before you install it.

What is a UGC video ad?

An ad that looks like a real person filmed it on their phone: someone talking to camera, holding the product, in a real room. The skill builds one with an AI creator and a five-shot direct-response structure: a hook, the problem, the product discovery, the result and a call to action.

What do I have to give the agent?

A brand that is already onboarded in Advibly and one angle: testimonial, car or on-the-go, unboxing, lifestyle demo or problem-solution. The product photo comes from your store catalog or asset library. If you do not describe a creator, the skill picks an archetype that fits your product.

How long does one ad take?

Most of the time is yours: you approve the script and then the storyboard. After that the five clips render in parallel, a few minutes each, and the finished vertical ad of about 40 seconds lands in your Advibly library with a link to fine-tune it in the video editor.

What does it cost?

The skill itself is free and open source. Running it spends your existing Advibly credits, roughly six images and five 8-second clips per ad, at no extra skill cost. Every generation shows its credit cost before it runs.

Which agents can run it?

Any agent that supports installable skills and the Model Context Protocol: Claude Code, Claude, Cursor, Codex, VS Code, OpenClaw, Hermes and ChatGPT. Install with one npx command, then connect the Advibly MCP so the skill has tools to call.

Does the creator sound the same in every shot?

Close, but not identical. Each clip synthesizes its own voice, so it drifts slightly between shots while the face and outfit stay locked. If you need one continuous voice, the skill points you to Advibly’s talking-actor tool instead.