The UGC Ad Template: Hook, Body, CTA

Almost every UGC ad that performs uses the same three-part skeleton: a hook that stops the scroll in the first three seconds, a body that earns the next fifteen, and a CTA that tells the viewer exactly what to do. The reason it repeats is that it maps to how attention actually decays on a feed. This is that template with real timing, three filled examples for different products, and the part most guides skip: how to turn each beat into an AI generation prompt so you can produce and re-test the structure at volume.

TL;DR

• 0 to 3 seconds is the hook. One idea, one visual, the highest-tension moment. If the hook fails, nothing after it matters.
• 3 to 20 seconds is the body. Three beats: problem, product-in-use, proof. This is where the sale is built.
• 20 to 30 seconds is the CTA. Name the action, add a reason to act now, show the product or offer on screen.
• The whole thing is promptable. Each beat becomes a 5-second clip you generate on 8frame at $0.28 to $0.40 per Kling 3.0 clip, then assemble in your editor.

Why the 3-part structure works

UGC ads live or die on retention. The feed gives you three seconds to earn the next three, and platforms reward the ads that hold attention with cheaper distribution. The hook-body-CTA structure exists because it front-loads tension (the hook), pays it off with a reason to believe (the body), and closes before the viewer's attention runs out (the CTA). Ads that wander, or that put the CTA at second 45 of a 60-second clip, leak viewers at every beat.

The structure is also what makes UGC testable at scale. Because the beats are modular, you can hold the body and CTA constant and swap only the hook, which is where the largest performance variance lives. That is the whole premise of UGC hook formulas for AI ads: the hook is the variable, the rest is the frame. If you are new to the format itself, what is UGC covers the definition and why brands lean on it.

The template, beat by beat

The hook (0 to 3 seconds)

One idea. The hook is a single high-tension moment, not a setup. The strongest hooks land on: a visible problem ("my roots grew out in four days"), a bold claim ("I stopped buying three products and use this one"), a pattern interrupt (an unexpected visual or sound), or social proof ("this sold out twice this month").

The camera should be close and the energy immediate. No logo intro, no slow pan into the scene. You are buying the next three seconds, nothing more.

The body (3 to 20 seconds)

Three sub-beats, roughly five seconds each:

• Problem (3 to 8s): name the pain the viewer recognizes. This is the "here's why the hook matters" beat.
• Product-in-use (8 to 15s): show the product solving it. Hands, texture, the actual moment of use. This is the most important body beat because it is the proof the claim is real.
• Proof (15 to 20s): the result, a before-and-after, or a one-line credibility marker ("dermatologist-tested," "10,000 five-star reviews").

The body is where you earn the sale. If the hook is a promise, the body is the evidence.

The CTA (20 to 30 seconds)

Three elements: the action ("tap the link," "get yours"), the reason to act now (offer, scarcity, or restock), and the product or offer visible on screen. Keep it short. The viewer who reaches second 20 is warm; do not talk them back out of it.

For the language that fills each beat, the UGC script templates for AI video give you fill-in-the-blank copy that maps directly onto this structure.

Three filled examples

Example 1: Skincare serum

• Hook (0-3s): Close-up on face, finger pointing at a visible blemish. "This cleared in three days and I'm still shocked."
• Problem (3-8s): "I'd tried every acid, every spot treatment, nothing held."
• Product-in-use (8-15s): Hand dispensing serum, applying to the area, texture visible on skin.
• Proof (15-20s): Split screen, day one versus day three. "Same lighting, no filter."
• CTA (20-30s): Bottle on screen, "20% off this week only. Link in bio."

Example 2: Cordless vacuum

• Hook (0-3s): Vacuum lifting an alarming amount of pet hair off a rug in one pass. "I have never seen my carpet this clean."
• Problem (3-8s): "Two cats, one shedding season, and a vacuum that gave up."
• Product-in-use (8-15s): In-hand, cordless, moving across floor and up onto a couch.
• Proof (15-20s): The full dust bin held up to camera. "One room. That's it."
• CTA (20-30s): Product on screen with price. "Under $200 and it's back in stock. Get yours."

Example 3: Protein snack bar

• Hook (0-3s): Bite, close on the cross-section. "Twenty grams of protein that doesn't taste like cardboard."
• Problem (3-8s): "Every protein bar I bought sat in the drawer because they're chalky."
• Product-in-use (8-15s): Unwrapping, eating on the go, in a gym bag.
• Proof (15-20s): Nutrition panel on screen. "Twenty grams, four ingredients, no weird aftertaste."
• CTA (20-30s): Multipack on screen. "First box is 15% off. Tap to try it."

Turning each beat into a prompt

This is where the template becomes production. Each beat is a 5-second clip you generate, then assemble into the 30-second ad. On 8frame, the pattern is:

Lock the person once. If your ad has a talking presenter across beats, lock the identity with Higgsfield Soul 2.0 from a reference image so the same face carries all beats. 8frame has no library of ready-made avatars; character continuity comes from identity-locking a reference, which is why the same reference feeds every clip. See what is identity locking in AI for the mechanics.

Prompt each beat as a shot. Structure: subject, action, framing, lighting. Example prompts mapped to the skincare ad above:

Hook: Close-up of a woman's face, natural window light, finger indicating cheek. Handheld phone-camera aesthetic, slightly imperfect. 9:16.

Product-in-use: Hand dispensing a drop of clear serum from a glass dropper, applying to cheek, macro focus on skin texture. Soft daylight. 9:16.

Proof: Split-screen composition, same face left and right, identical framing and lighting, subtle skin difference. 9:16.

Generate each beat with Kling 3.0 at 5 seconds, 1080p, 9:16, for $0.28 to $0.40 per clip. Use Nano Banana Pro at $0.04 to $0.08 for any product still you need as a reference or CTA card. A full 30-second ad is five to six clips, so roughly $2 to $3 in compute per finished ad before you assemble it.

The payoff of promptable beats is variant volume. Hold the body and CTA prompts constant, rewrite only the hook prompt, and you generate ten hook variants of the same ad for a few dollars. That set is exactly what the AI ad variant testing workflow is built to find winners from.

Common mistakes

Hook is a setup, not a moment. "Hi guys, so today I want to talk about" is not a hook, it is throat-clearing. Start on the highest-tension frame.

Body with no product-in-use beat. If the viewer never sees the product actually working, the claim is unproven. The in-use beat is non-negotiable.

CTA with no reason to act now. "Check it out" gives the warm viewer permission to leave and come back never. Add the offer or the scarcity.

Over-length. Thirty seconds is the ceiling for most feed placements. If your body runs long, cut the proof beat before you cut product-in-use.

FAQ

How long should a UGC ad be?

For feed and Reels placements, 15 to 30 seconds is the working range, with 30 as a practical ceiling. Shorter cuts (15 seconds: hook, one body beat, CTA) often win on cost-per-result because they finish before attention decays. Generate the full 30-second version and cut a 15-second variant from the same clips to test both.

Can I make a UGC ad without a real creator on camera?

Yes. Lock a character identity from a reference image with Higgsfield Soul 2.0 and generate each beat as an AI clip. Be aware that 8frame has no ready-made avatar library, so continuity depends on feeding the same identity reference to every beat. If you need a ready avatar reading a script quickly, an avatar-actor tool may be the faster path; if you need product-in-hand shots and full scene control, the canvas route wins.

How do I test which hook works best?

Keep the body and CTA prompts constant and rewrite only the hook prompt to produce 8 to 12 hook variants of the same ad. Launch them grouped by hook type at low daily spend, give each 72 hours to accumulate hook retention data, then scale the two or three that separate from the field. The hook is where the largest performance variance lives, so that is where your test budget should go.

---

Every beat in this template is a clip you can generate on the 8frame canvas from one product reference. Build the 30-second master once, then swap hooks to spin out a full test set for a few dollars. Pair it with the product video template when the ad needs to lead with the product instead of a presenter.