AI Creative OS
01Make the two references first02Lock the hero frame03Write the prompt as a shot list04Run it through one connector05QC like an editor, then package the breakdown

The UGC Prompt System

How the VELA serum ad was made with no creator and no product shoot: two reference images, one hero frame, one timed prompt, one video model, all through one connector. The ready-made prompt lives inside AI Creative OS.
What you'll make
Want this running for you?Build it step by step inside AI Creative OS, with me in the thread.Get Your First Agent →
01Step 1
Make the two references first

Open Nano Banana Pro (gemini-3-pro-image) and make two stills, not a video. One: the product on a seamless white background, label facing the lens, with the label text in quotes in your prompt so it renders exactly.

Two: a phone selfie of the creator in the room where the ad will happen (a vanity, a kitchen, a car). Write the person like a casting note: age, hair, freckles, earrings, sweatshirt.

Add the realism layer every time:

  • visible pores
  • flyaway hairs
  • no smoothing
  • subtle sensor noise
  • muted natural grade

For a client, skip step one and use their real packshot. Fictional brands only when you are practising.

02Step 2
Lock the hero frame

Make one more image with both references attached and say what each one is for: reference 1 is the product, reproduce it exactly; reference 2 is the person and the room, identity only. Describe the first frame of the ad: her at the vanity leaning toward a stationary phone, the bottle on the table with the label toward camera.

This frame is the start frame of the video, so face and product are locked before any motion exists. If the bottle or the face is off here, fix it here.

Stills cost cents, videos cost dollars.

03Step 3
Write the prompt as a shot list

Keep the shape and change only the product, the person and the four dialogue lines.

Keep the shape:

  • an identity lock at the top
  • then four timed shots (hook, product reveal, application, result)
  • each with one camera position
  • one action and one line in quotes
  • then a style line
  • a negative line

Stay under 300 words. A video model treats a shot list as an edit; it treats a paragraph of adjectives as a mood.

Want this running for you?Build it step by step inside AI Creative OS, with me in the thread.Get Your First Agent →
04Step 4
Run it through one connector

In Claude Code, the ElevenLabs connector gives you the image model and the video model behind one login (add it with claude mcp add --transport http elevenlabs https://api.elevenlabs.io/v1/mcp). Wire the hero frame into Kling 3 Pro as the start frame, 15 seconds, 16:9, native audio on, and ask for the estimate before the run.

My run: every still including the variations came to about $2.30 in credits and the 15-second video to about $3, so the whole ad kit was under $6. The video is the expensive step, so estimate it first and run it once.

Seedance 2.5 is the alternative when you want reference images on the video call itself.

05Step 5
QC like an editor, then package the breakdown

Watch the clip three times: once for the face across the cuts, once for the bottle (same label, same cap, no morphing), once for the hands. If one shot drifts, regenerate the still for that shot and re-run, never re-roll blind.

Then post the breakdown, not just the ad: the clip on top, the two references under it, and the prompt scrolling while it plays. That format is the post you are reading this from, and it is what stops the scroll.

🔒 UGC PROMPT SYSTEM pack · members
The exact UGC video prompt

This guide is the full process. The ready-made version lives inside AI Creative OS on Skool, next to every other prompt pack I post.

  • The exact timed video prompt behind the serum ad
  • Where to change the product, the person and the four dialogue lines
Start free trial →
There's a free trial, so you can open the pack today.
Not ready for that?
Drop your email and 3 more guides open in the Library, free.