A convincing AI UGC product demo should show one clear action, not merely place a product beside a talking avatar. The viewer needs to understand what the product is, how it is used and what happens next.
The most reliable approach is to build the ad as short, controlled shots. Create accurate still frames first, animate one action per clip, then add dialogue, captions and sound. This gives you more control than asking a video model to invent an entire demonstration in one generation.
Choose one action the viewer can verify
Start with the most visual benefit. A skincare bottle can dispense product onto a fingertip. A bag can open to reveal its compartments. A kitchen product can be assembled, poured or served.
One ad should answer one practical question:
- How is the product opened or applied?
- What size is it in a person's hand?
- What is included in the package?
- What does the texture or finish look like?
- What problem does one feature address?
Write the action in one sentence before scripting. If that is difficult, the demonstration is probably too complicated for one short ad.
Plan the demo as five separate shots
A useful structure is:
- Hook: Show the product or result immediately.
- Context: Establish the creator and problem.
- Action: Demonstrate one physical step.
- Proof: Cut closer to the visible detail.
- Close: Give a clear next step.
This follows the hook, body and close structure in TikTok's current Creative Codes guidance. TikTok also recommends vertical 9:16 footage, high-resolution source material and leaving room for its interface.
Turn the structure into a shot list:
Shot 1: Hand holds the bottle close to camera. Shot 2: Creator explains the application step. Shot 3: One pump dispenses onto the hand. Shot 4: Close-up of the product texture. Shot 5: Product on the shelf with CTA text.
This exposes difficult moments early. If the pump action or label cannot be shown accurately, redesign that shot before generating the rest.
Prepare creator and product references
Use a clear creator reference for identity and several clean product images for packaging. A practical product set includes a front view, a three-quarter view and any angle needed for the action, such as the open lid or applicator.
Do not use a lifestyle photo as your only product reference. Reflections, fingers and perspective can hide the exact shape. A clean packshot is easier for a model to interpret.
Record the details that must remain fixed:
- Product shape, proportions and colour
- Cap, pump, handle or closure
- Logo placement and label layout
- Number and position of components
- Creator face, hands and proportions
The guide to stopping AI UGC from changing your product covers a deeper product-accuracy workflow. For a demo, match the reference angle to the action. A front packshot will not fully explain how an applicator looks from above.
Build and approve still frames first
Create one still for each planned shot. Do not animate until the product, hand position, framing and creator identity are acceptable.
A setup prompt might read:
Vertical creator-style phone video frame in a bright bathroom. The same recurring creator holds the referenced skincare bottle beside her face, label facing the camera. Preserve the exact bottle shape, pump, colours and label layout. Natural window light, chest-up framing, realistic skin, room for captions at the top.
For the action:
Close view of the creator's hands. One hand holds the referenced bottle upright while the other is beneath the pump. Preserve the exact packaging, closure, logo placement and hand anatomy. Neutral bathroom counter, realistic phone-camera look.
For proof:
Macro product demonstration frame showing a small amount of translucent gel on the back of the creator's hand. The referenced bottle remains beside it with its shape and colours unchanged. Sharp focus on texture, soft background.
Check each image at full size for duplicate fingers, fused contact points, floating packaging, changed labels and impossible mechanisms. Use local repairs where possible. The guides to fixing AI-generated hands and preserving logos and labels cover those problems in detail.
Animate one action per clip
Use an approved still as the first frame for image-to-video. Let the image define the composition and product appearance, then use the video prompt mainly for motion.
Runway's current image-to-video prompting guide recommends this division. It says the input image establishes subject matter, composition, lighting and style, while the text prompt should focus on motion. It also warns that source-image artifacts may intensify during animation.
Keep each prompt narrow:
The thumb presses the pump down once. A small amount of gel dispenses onto the other hand. The camera remains still.
The creator slowly rotates the bottle toward the camera until the front label is visible. Subtle handheld phone movement.
The creator spreads the gel across the back of her hand in one smooth motion. The camera gently moves closer.
Do not combine opening, pouring, applying, reacting and speaking in one clip. Complex hand-to-product interactions are already difficult. Several simultaneous actions make failures harder to diagnose.
Use a hybrid shot when precision matters
Small printed instructions, flexible packaging, liquid pouring, jewellery clasps and mechanical parts can change from frame to frame. Use real footage for the precision-critical close-up when needed. A phone clip of genuine hands operating the product can sit between AI creator shots. Match the crop, light direction and colour grade.
You can also imply a difficult action with editing. Cut from the closed package to the correctly opened product, then use an accurate sound effect to bridge the change.
Do not turn a demonstration into a fake testimonial
If viewers can see the product dispensing, dialogue can explain why the step matters without pretending the AI presenter personally tested it.
Use supplied, supportable information:
"Use one pump and spread it evenly across clean skin."
Avoid invented experiences such as "I used this for a week and it cleared my skin" unless the words accurately represent a real, documented testimonial and their use is permitted.
The US Federal Trade Commission says in its guidance on AI avatars and testimonials that AI avatars are not banned in marketing, but false underlying testimonials can be prohibited and avatar use can still be deceptive. Apply the rules for every market and platform where the ad will run, and make required disclosures easy to notice.
Edit for clarity
Put the product on screen early, trim dead air and use close-ups when they clarify the action. Captions should reinforce the demo, not cover the product or hands. Keep important text away from the platform interface.
Watch once without sound. The action should still make sense. Then listen without watching. The voiceover should remain accurate and should not promise something the visuals do not show.
Approve the finished demo
Review the export frame by frame:
- Is the product the correct variant throughout?
- Does the packaging mechanism work realistically?
- Are hands and contact points believable?
- Does each clip contain one clear action?
- Is the product visible during the main claim?
- Are captions readable and inside safe areas?
- Are all claims supportable?
- Is any required disclosure clear?
In Rasgo AI, save the approved creator and product references before producing variations. Change the hook, setting or CTA only after the core demonstration works. A repeatable demo is more useful than an impressive clip that shows the wrong product action.
Explore AI video generator for the workflow, then create a video in Rasgo.
Create your next Rasgo visual
Turn the ideas from this guide into generated images, creator content, product shots, or video-ready concepts.