The UGC Ad Creative Playbook
The whole method, no gate. Hooks, script structure, actor consistency, realism direction, and second-by-second scripting, pulled from real ad teardowns. Copy any block and swap the brackets. Or skip all of it: our storyboard-first tool runs this exact system for you, no prompting.
Section 1
Hooks that stop the scroll
71% of viewers decide whether to keep watching inside the first 3 seconds. Write the hook first, in isolation, before you touch the rest of the script. None of these open on the product. They open on the viewer. Swap the brackets for your product and read it out loud, if it does not create curiosity or name a specific pain, rewrite it.
Hook 1: Curiosity gap
Promises a payoff without giving it away, so the viewer stays to find out.
I found the one thing that actually fixed my [specific problem], and it was not what I expected.
[Then cut straight to holding the product. Do not explain yet.]Hook 2: Bold claim / hot take
Challenges what the viewer already does, which forces a reaction.
Stop wasting money on [the common solution everyone buys]. Here is what I switched to.Hook 3: Specific personal stat
A real number beats a vague question every time. Specificity is what makes UGC read as real.
I was spending $[amount] a month on [category] that was not working. Then I tried this.Hook 4: Result-first
Leads with the payoff. The product enters only after the hook has earned the next 3 seconds.
This is what [timeframe, e.g. two weeks] of using [product] did to my [visible result].Hook 5: Pattern-interrupt callout
Names a specific, personal pain the viewer feels in the first breath.
Nobody warns you about [specific frustration in the category]. I dealt with it for [timeframe] before I found [product].Section 2
The 4-block script skeleton
Every converting UGC script is four blocks on a clock: hook, problem, solution, CTA. Keep the whole thing to 9 to 15 seconds. Brief the hook and the key belief, then leave room for natural delivery. Over-scripting every word is the fastest way to make an ad feel like an ad.
The full script template
Drop any hook from Section 1 into the 0 to 3 slot and the structure holds.
HOOK (0-3s): [one line from Section 1, bold claim or specific stat]
PROBLEM (3-8s): [name one specific, personal pain. Not 'tired of bad skin?' but 'I was spending $200 a month on skincare that did nothing.']
SOLUTION / DEMO (8-20s): [show the product solving it. One clear benefit, 5 to 8 words. Not a feature list.]
CTA (final 3-5s): [one action, 2 to 3 words: 'Shop now', 'Get yours'. Destination matches the words, product page, never the homepage.]Section 3
Actor prompts: anchor the face before you render
AI actors drift because video models have no memory. Same description, different bone structure on every clip. The fix: lock one approved reference frame (a gray-background contact sheet), then force every downstream generation to match it. The average tool needs 15 to 20 retries per scene without an anchor, and every retry is a render you pay for.
Step 1: Generate the hero reference frame
This one approved image becomes your identity source of truth. Neutral background, slightly desaturated skin so the model does not swing warm outdoors and cool indoors.
A single clean portrait of [age] [gender] with [hair], [skin tone], [distinct feature], shot on a plain neutral gray background, even soft lighting, natural skin texture with visible pores, no makeup gloss, slightly desaturated skin tone, front-facing, relaxed expression, photographic, not stylized.Step 2: Pass the anchor into every clip
Attach the approved frame as the reference image on every scene, not just the first. Each clip copies the same face instead of inventing one.
Use the attached reference image as the exact identity for this shot. Keep the same face geometry, skin tone, and hair. Do not reinterpret the person. [Then describe only the new action, setting, and framing.]Step 3: Lock wardrobe, skin, and voice
Skin tone and clothing shade drift first. Describe wardrobe identically in every scene prompt, and pin one base audio sample as reference audio so the voice does not change between clips.
Wardrobe (identical every clip): [garment], [exact color], [material]. Keep lighting and background fixed within this video. Reference audio: [same base voice sample on every render]. Vary setting and wardrobe ONLY across different videos in the batch, never inside one.Section 4
Product-demo prompts
The demo shows the product solving the problem, one clear benefit, in the hands of the same anchored actor. Keep it single-actor and hand-held. One believable person talking to camera outperforms a crowded scene and sidesteps multi-character identity blur, which no major tool has solved.
First-use / unboxing demo
Shows the moment of use, not a spec sheet. The benefit lands in a glance.
Same actor (reference attached), holding [product] up to the camera at arm's length, natural window light, showing [the one specific action that proves the benefit]. Genuine reaction, not a presenter smile. Handheld, slight shake, phone-camera framing.Close-up benefit demo
A tight crop on the result is the proof. Let the visual carry the claim.
Close-up on [the visible result or texture the product changes], same lighting and setting as the previous clip, shallow phone-camera depth, no studio polish. Actor's hand enters frame to [interact with product]. Real, slightly imperfect, iPhone-shot look.Section 5
iPhone-UGC realism settings
The single thing that separates UGC that converts from AI slop that gets swiped: it looks shot on a phone by a real person, not rendered by a studio. UGC-style creative pulls a 31 to 41% higher hook rate than polished studio production. Append this settings block to every image and video prompt.
The realism settings block (append to every prompt)
Paste this at the end of any generation prompt to force the iPhone-shot look.
STYLE: shot on iPhone front camera, vertical 9:16, natural available light (window or overhead, never studio softbox), slight handheld shake, authentic skin texture with visible pores and minor imperfections, no color grading, no cinematic lens flare, no shallow-focus bokeh backdrop, plain everyday setting (bedroom, kitchen, car, bathroom mirror), casual framing slightly off-center. Read as a real person filming themselves, not an ad.Section 6
Timestamp prompting
Video models cap around 15 seconds per clip, and they average the whole shot unless you tell them what happens when. Prompt the clip second-by-second so the right thing shows at the right beat. This is how a storyboard becomes a scene instead of a blur.
Timestamped clip prompt
Maps the four script blocks onto a single 15-second clip's timeline.
15-second clip, same actor (reference attached), handheld iPhone look:
0-3s: actor looks into camera, delivers the hook line, [product] not yet visible.
3-8s: actor gestures while naming the problem, frustrated but real.
8-12s: actor lifts [product] into frame, demonstrates [benefit action].
12-15s: close-up on [result], actor delivers 2-word CTA.
Keep lighting, background, and wardrobe identical across the full 15 seconds.Section 7
The test loop that finds winners
One script is a coin flip. Only 1 to 3 of every 10 creatives become real winners, so a single script is a bet, not a plan. Write the hook as a variable and test many versions. The bottleneck is never the writing, it is the producing, so preview cheap and render only the survivors.
The batch variation setup
Same product, same problem, same offer, same actor. Only the first 3 seconds change, so the hook is the only variable.
Take [product] and one problem statement. Write 8 to 10 different hooks from Section 1. Build each into the 4-block script. Keep the SAME anchored actor, wardrobe, and setting across all of them. The opening line is the ONLY thing that changes between variations.The script-to-test loop
A weak hook dies for the price of a picture, not a full shoot.
1. Write 8-10 hooks (same product + problem).
2. Build each as a 4-block script on the 9-15s clock.
3. Preview each as a cheap storyboard image. Kill the weak ones on sight.
4. Render only the 4-6 survivors into video.
5. Launch, find the 1-2 winners, then write variations of THOSE next.Skip the copy-paste
Generate these instead of writing them
This pack is the manual version. HighQualityUGC runs the whole loop for you: it anchors one actor to a contact sheet, drafts each hook as a cheap storyboard you approve before any video credits spend, then renders only the survivors. Same method, none of the typing.