Maybe more. The good creators charge $800+.

And here's what you get for that money:

Want variations? Pay again. Different demographic? Pay again. A/B test hooks? You guessed it.

The economics are broken.

Last week, Kling 3.0 dropped. And buried in the announcement was a feature that changes everything.

It's called "Elements."

Upload a reference image of a person. The AI locks their identity across every video you generate. But here's the part most people miss: Elements isn't just for faces. Upload your SaaS dashboard, your app screen, your product packaging (and it appears in the video as a locked element too). Your AI creator actually using your actual product.

Combine that with native lip-sync in 5 languages, multi-shot storyboards (up to 6 shots), and 4K output...

You're not generating videos anymore. You're building a creator roster with your real product baked in.

I’m giving away a Skill you can use for free to make your UGC army. Here's exactly how I'm using it.

Founders spend too much time drafting the same kinds of messages. Wispr Flow turns spoken thinking into final-draft writing so you can record investor updates, product briefs, and run-of-the-mill status notes by voice. Use saved snippets for recurring intros, insert calendar links by voice, and keep comms consistent across the team. It preserves your tone, fixes punctuation, and formats lists so you send confident messages fast. Works on Mac, Windows, and iPhone. Try Wispr Flow for founders.

Start flowing free

Before Elements, AI video had a fatal flaw for UGC: consistency.

You'd generate a testimonial video. Great. Generate another with the same "person"? Completely different face. Different vibe. Different everything.

No brand would use that. UGC works because people recognize the creator. Trust compounds across videos.

Elements solves this. And it goes further than most people realize.

Upload one reference image of a person. The model creates an identity lock. Now that exact person (same face, same features, same vibe) appears in every video you generate.

But Elements supports multiple locked references per video. Upload your creator and your product (a SaaS dashboard screenshot, an app screen, product packaging) and both appear consistently. Your AI creator scrolling through your actual app. Holding your actual product. Reacting to your actual interface.

Same product. Different scripts. Different settings.

That's the foundation of a UGC factory.

Here's the architecture I've been testing. The key insight: visuals before script. Design how each shot looks and moves first, then write dialogue that fits. The script serves the visuals, not the other way around.

Let me break down each step.

First, you need characters that match your target audience.

(As always) I’m using Nano Banana Pro (Google's image model via Replicate) to generate reference images.

The secret: Most AI image prompts create obviously fake people. The fix is adding imperfections deliberately.

The Photorealistic Prompt Formula:

Example for a DTC skincare brand:

Generate 5 variations. Pick the best. That's your creator.

The key: Natural, not perfect. Add imperfections deliberately. Flyaway hairs, under-eye texture, asymmetry. The negative prompt does half the work by telling the AI what not to do.

Once you have your reference, you pass it to the Elements system via fal.ai's API. Now every video you generate uses that locked identity.

I built a roster of 6 characters:

Different products call different creators from the roster.

Key principle: visuals before script. Design how each shot looks and moves first, then write dialogue to fit. Most people do this backward—they write a script and try to make visuals match. That leads to talking-head videos where nothing changes visually.

This is where Kling 3.0's multi-shot feature shines. You can define up to 6 distinct shots in one generation. The AI handles transitions, shot composition, and character continuity automatically.

Variable shot durations: Each shot can be 3s, 4s, or 5s. Hooks should be punchy (3s). Demos need room to breathe (5s). Total must sum to your target duration.

Here's my testimonial storyboard template:

Notice: no dialogue yet. Just pure visual design. The script comes next, fitted to these durations.

The AI generates all 4 shots with smooth transitions. Same character throughout.

15 seconds of UGC in one generation.

The 2-of-4 Rule: At least 2 of your 4 shots must have a meaningful position or angle change. If every shot is "sitting at counter, talking" → transitions feel weird and there's no visual progression.

Build variety into every storyboard:

Now that your visual storyboard is locked, write dialogue that fits each shot's duration.

The word budget: ~2.5 words per second. A 3-second shot gets ~7 words. A 5-second shot gets ~12. Go over and the lip-sync feels rushed. Go under and there's dead air.

Generic AI copy = obvious fake. Here's what actually sounds human:

What makes it work:

The pattern: Specific details + filler words + natural rhythm. Map every line to a shot from your beat sheet. The script serves the visuals, not the other way around.

Most AI-generated UGC gets rejected by viewers in the first 2 seconds. Human brains detect unnatural movement.

We've spent our entire lives watching real people move. Our pattern recognition is insanely sophisticated. When movement is wrong, even slightly, we feel it immediately.

"Something's weird about this."
"This doesn't look right."
"Fake."

The issue: Most people prompt AI video like they prompt images. They describe what's in frame, not how it moves.

"Woman talking about skincare."

Kling generates motion. Your prompt needs to describe movement patterns that match what humans unconsciously expect.

Here's what human brains expect from real UGC:

What makes human brains reject it:

The fix is prompting for the imperfections humans expect.

The 6 Elements of a Strong Prompt:

The 4 Rules:

Weak vs Strong:

Write prompts like single-shot film directions. One flowing sentence, not a keyword list.

Your hook needs to happen in the first second. Not after buildup.

Most people focus on the script. But UGC hooks work in layers:

The Best performing UGC uses 2-3 layers simultaneously.

Here's the hook formula that actually works:

Discovery Hooks (High Performers):

Disbelief Hooks:

Pattern Interrupt Hooks:

The pattern: Audio hook + visual reaction + optional text overlay.

Your AI character says the hook while their face shows genuine surprise or excitement. Text reinforces with different words. Three layers, one second.

Here's what nobody talks about: your audience has been trained since
birth to detect fake human behavior.

Not consciously. Unconsciously.

When audio is too clean, too perfect, too "produced", people don't
think "that's an AI ad." They just feel something's off. And they scroll.

I tested this with a client last week. I generated 10 testimonial
videos. Clean audio on 5, imperfect on the other 5.

The clean videos: 0.6% CTR
The audio: 2.8% CTR

People just didn't engage.

Here's what authentic UGC audio actually sounds like (and why your
brain trusts it):

The algorithm doesn't judge. But your audience's unconscious pattern
recognition? That's been training for 200,000 years.

Here's the uncomfortable truth: most AI-generated UGC fails not because the video quality is bad, but because it smells like an ad.

Before generating, I run every concept through 4 mandatory checks. They're about matching what humans unconsciously expect from authentic content.

The 4 HUMAN DETECTOR CHECKS:

If any test fails humans scroll. Meta sees low engagement. The ad dies.

You're not gaming the system. You're creating content that passes the
test humans have been running unconsciously since childhood.

Now we’re going to flip the economics.

At $500 per real UGC video, that's $7,500 in creative testing.

With this system: maybe $50 in API costs.

We run paid social for DTC brands. UGC is always the bottleneck.

We're not replacing human UGC creators. We're filling the gap between "we need 10 creative variants" and "our budget is 2 videos."

Here's the smart play: AI-generated UGC as creative testing before paying real creators.

Generate 10 AI testimonials with different hooks, angles, demographics. Run them at low spend. See what resonates.

The winner? Now brief a real creator to recreate that winning combo.

You're not guessing anymore. You're validating with data before writing the bigger check.

Is this perfect? No. Here's what doesn't work yet:

Lip Sync Drift: Once you hit 10 seconds the lip sync starts to drift. Ask the skill to generate in smaller chunks if you find this.

Zero editing: You'll still want to add captions, music, and a cut or two. This isn't "press button, receive finished ad."

Generation artifacts: Sometimes you get weird stuff, object morphing (phone turns into product), extra hands appearing, face drift mid-video. It's stochastic failure. Just reroll. Expect around 4 regenerations per final video.

Scene transitions: If your shots are too similar (sitting → still sitting → still sitting), transitions feel awkward. Build visual variety into your storyboard. Position changes, angle changes, framing changes.

But for the core job, consistent character + natural testimonial + multiple variants, it's shockingly good.

This definitely DOES NOT replace your best-performing human creators. What it replaces is the $15,000/month you're spending on "good enough" volume.

Six months ago, AI video was a novelty, unreliable. Interesting demos, unusable for real work.

But that gap between "cool tech demo" and "wait this might be actually useful for my business" is closing fast.

But now that we have character persistence, product consistency, and realistic motion, that changes the calculus. The power here isn’t creating one-off videos. Think of this running across dozens of stealth accounts, finding the best hooks, angles, strategies all while getting eyeballs, and then giving your human creators the best of the best. That's a content strategy.

The brands that figure this out first will have 10x the creative testing velocity at 1/10th the cost.

The UGC arbitrage window is open. It won't stay open forever.

I'm giving away the entire UGC Content Factory skill. The same system I use above, automated.

DOWNLOAD THE FULL SKILL

You get:

Takes about 10 minutes to set up. Then you're generating.

If you'd rather skip the learning curve and just get the content, we're opening a few UGC strategy sessions this month.

We'll build your first character roster, generate test videos for your product live, and map out a variant testing plan so you know exactly which hooks and creators to scale.

Book a UGC Strategy Session →

Reply and tell me: what product would you test first with this system? I read every response.

Let's build,

Matt

P.S. The irony isn't lost on me that I spent years building relationships with great UGC creators... and now I'm teaching you to replace some of that work with AI. The difference? The best human creators are still irreplaceable. What's replaceable is the $400 "good enough" middle tier. AI just became the new floor.

🎁 Get rewarded for sharing! My team grew NEW accounts to over 50 million views in just a few months. We made an AI viral hook generator so you can follow the same hook frameworks that we do.

Invite your friends to join, and when they sign up, you’ll unlock our AI Viral Hook Generator—the ultimate tool for creating scroll-stopping content.

{{rp_personalized_text}}

Copy and paste this link: {{rp_refer_url}}