Track map
Detailed content
🤖 The AI workflow
Three tools working together: one that thinks and coordinates, one that generates the video, and a platform that brings it all together. You provide the template and get the ad.
An AI “conductor” (like Claude Code) that receives your request, analyzes the reference video, and gives instructions to the other tools.
It’s what turns “I want an ad like this” into concrete steps, without you having to do everything by hand.
Coordinates · analyzes · gives orders. The brain of the workflow.
The tool (such as Seedance 2.0) that actually creates the UGC-style video from an image and instructions.
It’s what works the “miracle” of turning a product photo into a video of someone talking.
Image + instruction → video. The production arm.
A platform (like Arcads) with a library of actors, voices, and everything ready to go, where you build the ad without code.
It’s the easiest way for people who don’t code: you drag, choose, and generate.
Ready-made actors + voices · no code · simple interface.
The complete sequence: you provide a sample video, AI analyzes it, writes a script in the same style for your product, and generates the video.
It’s the heart of Track 4. Understanding the workflow puts you in control of every step.
Model → analysis → script → video. Four steps.
You can also take a video that works and ask: "replace the product with this one" and "replace the actor with this other one".
It’s the fastest way to adapt a winner to your product while keeping the winning formula.
Same structure, your product, your actor. Smart adaptation.
You provide the Track 3 template + a photo of your product, and the AI writes your ad script in that style.
It’s the bridge between studying (Track 3) and the finished video (this module).
Template + product photo → finished script.
🎙️ Actor and Voice: The Voice Is Everything
The golden rule of UGC: people forgive bad visuals, but they never forgive a bad voice. Learn to choose the right actor and engineer a voice that sounds human.
Platforms have libraries of AI “actors.” You search by gender, age, and type, then choose who will speak.
The right actor creates an immediate connection with the viewer.
Search by gender/age · young, adult, senior · choose and you’re done.
Choose an actor who resembles your ideal customer — the same age range, the same "world."
We trust people who are like us more. Different actors speak to different audiences.
Mirror your customer · senior audiences respond to seniors.
The most important takeaway from the learning path: the voice carries trust. People forgive poor visuals, but they don't forgive a bad voice.
If you can only perfect one thing, perfect the voice.
Voice = trust. Invest time here.
Instead of having AI "read" a text, you record yourself speaking, and the tool copies your cadence and delivers it in the avatar's voice.
A voice copied from real people sounds much more natural than a voice "reading."
Record your speech · AI copies the cadence · human result.
Two controls: stability (more steady vs. more expressive) and similarity (how much the voice resembles your recording).
Adjusting these settings is what fine-tunes the voice until it sounds truly natural.
Stability = consistency. Similarity = resemblance. Test both.
Before generating the entire video, listen to the voice preview and adjust the speed and controls.
Fixing the voice in the preview saves time and prevents you from generating a video with a bad voice.
Listen · adjust · only then generate the video.
🛡️ Realism, variations, and security
The details that separate an ad that sells from one that looks fake—and how to test several versions and protect your passwords along the way.
For UGC, don’t use the highest resolution possible. Everyday people don’t record in perfect 4K—that gives away that it’s an ad.
A little imperfection builds trust. Realism sells better than perfection.
Avoid 4K · make it look homemade · imperfections = authenticity.
Generate multiple versions of the same ad (different hooks, actors, and voices) and run them all to see which sells more.
It’s the same secret as Track 1: the winner emerges when you test several variations.
Many versions · let the audience choose · scale the winner.
Set up the video in the right format: vertical (9:16) for feed/stories and short (around 10-15 seconds).
The wrong format kills reach. Vertical and short is what the feed rewards.
9:16 · short and direct · designed for mobile.
The tools use “API keys” (an access password). Never paste that password directly into an AI conversation.
Ask the AI for a command that stores the password securely, outside the conversation history.
Key = password · never in chat · use a secure command.
Platforms charge by credit or monthly plan. You can start small and grow as results come in.
Knowing the cost helps you avoid surprises and calculate whether the ad is making a profit.
Credits or monthly subscription · start small · measure the return.
Run your variations, see which one sells, put more money behind the winner, and repeat the process.
It’s how the whole method pays for itself: you only scale what you’ve already proven works.
Test → measure → scale the winner → repeat.