AI video & content
AI-generated UGC for business: what our DAYFORM demo taught us
A convincing presenter is only one part of a useful product video. For our DAYFORM concept, we wanted a simple shot: a person in a sunlit kitchen gently lifting a cream bottle with a mint cap. Comparing two generated clips showed why the product, the spoken message, and the ending need their own acceptance checks.
The brief: one presenter, one product, one action
DAYFORM is a fictional bottle brand we used for an internal creative exercise. It was not client work, and no sales campaign or conversion study was attached to it. The brief specified chest-up smartphone framing, subtle handheld movement, an oatmeal shirt, natural-looking skin, and a readable charcoal wordmark on a cream bottle.
The action was deliberately small: look at the camera and gently lift the bottle. Even this short sequence gives a reviewer several things to check. Does the cap stay the same shape? Can a viewer identify the product? Does the hand obscure the label? Is the product still visible when the clip ends?
What we could actually see in the two outputs
We reviewed a four-second export labelled Veo 3.1 and a six-second export labelled Gemini Omni Flash in their filenames. Our review used sampled frames and full-resolution stills. The second clip’s exact prompt and settings were not independently verified, so this was a comparison of two outputs, not a controlled ranking of the models.
The first output had a close, casual selfie composition and a readable label in earlier frames. Its direct sunlight made strong shadows, the shirt looked warmer and darker than the brief, and the bottle moved out of view near the end. The second output had softer light and an oatmeal shirt closer to the requested styling; the bottle remained visible in the ending samples. Its label softened during the lift and became clearer in settled frames.
We preferred the second output for this concept because of the product visibility and visual match. That preference does not establish better model performance in general. We did not review full-motion playback or listen to these clips in this assessment, so we cannot claim that their speech, lip sync, or motion was ready for an advertisement.
Write the spoken line before generating the final take
Our supplied visual prompt did not specify dialogue. Both files contained audio tracks and apparent speaking mouth shapes, but that does not tell us what was said or whether it suits the brand. A polished face cannot compensate for a missing message.
For a short product concept, an illustrative line might be: “Meet DAYFORM. A little colour for your everyday routine.” This is a proposed script, not a transcript of the clips. Read the line aloud at the intended pace, leave room for the product gesture, and shorten the copy if it feels rushed. Keep any specifications, material claims, or benefits tied to facts the business can substantiate.
- Define the opening message and the one action you want the viewer to take.
- Specify language, delivery, pronunciation, and whether the presenter or a separate voiceover speaks.
- Decide where captions and the final call to action belong before choosing the framing.
- Keep the approved product details and presenter reference consistent between scenes.
Use a review sheet that separates appearance from readiness
For the next production, we would review the full clip at normal speed with sound, then revisit individual frames. A still-frame check is useful for detail; playback is necessary for pacing, movement, and speech. This worksheet makes the decision concrete.
| Check | What to inspect | When to revise |
|---|---|---|
| Product | Shape, cap, colours, label, hands, opening and ending visibility | Identity changes or the product disappears during the key message |
| Presenter | Face, clothing, eye line, gestures, continuity between scenes | Distracting changes or gestures that contradict the script |
| Voice | Exact words, pronunciation, lip sync, pauses, background sound | Missing claims review, wrong dialogue, clipped words, or timing mismatch |
| Edit | Captions, mobile legibility, platform interface overlap, final action | Unreadable text or an ending that gives no clear next step |
Describe the format honestly
UGC usually refers to user-generated content. Here, “UGC-style” describes the direct-to-camera visual format; the presenter is synthetic. Do not turn that into a claim that a real customer bought, used, or recommended a product. A business can present an AI spokesperson explaining an offer without inventing a personal experience.
Review the destination platform’s disclosure settings before uploading. YouTube’s guidance requires disclosure for qualifying realistic altered or synthetic content. A realistic AI presenter should prompt that review. The platform guidance is a publishing check, not evidence that a particular advertisement will be approved or perform well.
References: YouTube: disclosing altered or synthetic content
Start with one concept you can evaluate
A useful first brief contains the actual offer, intended audience, approved product references, one short message, and one destination for interested viewers. Agree what the final deliverable includes: script, generated scenes, voice, captions, revisions, and export formats. Review those items together before generating a batch of variants.
Our DAYFORM exercise helped us identify production decisions, not a formula for sales. For a real business, the next step is a reviewed video connected to a clear enquiry journey, followed by measurement of relevant responses. Attractive output is a starting point; it still needs an accurate offer and a working way to contact the business.
Sources & further reading
Service guidance can change; check the linked documentation when implementing.