AI content · September 2026
UGC AI is what all the big companies are looking for
A client asked what we could do in an afternoon. We answered with one AI-generated still, one AI-generated UGC video, and a template you can reuse for your own product. Here is every image, every prompt, and the reason each choice was made.
Every brand is chasing the same thing on TikTok and Instagram: content that looks like a real person filmed it in their bathroom, not a studio. That format — UGC, user-generated content — is what converts, and it is expensive to produce at volume because it needs a creator, a location, a product in hand and a day of shooting per variation.
A client asked us what we could do in a short period of time. This is what we sent back: one composite still, generated from three reference images, turned into a talking-head UGC video (and, as an extra, a hook clip). No shoot, no creator, no location. Below is the whole recipe, with the exact inputs and prompts, and a template builder so you can swap in your own product.
The result
The real video is what we deliver; the hook is an optional extra for the first second of a TikTok or Instagram cut. Both come from the same still.
The recipe in one line
Three references → one composite still → one video.
3 references
1 still
1 video
Step 0 — The three inputs
Character sheet — four views
Garment — a flat
Product — the packaging render
The character. A four-view model sheet. Ours came from Pinterest; the four angles (front, three-quarter, profile, full body) are what keep the face consistent. If you do not have one, generate it — the template kit includes a character-sheet prompt (T0), and the same technique works with a fully synthetic character.
The garment. A flat of the T-shirt. A flat is better than a worn photo because the model sees the artwork without folds and reproduces it accurately.
The product. The packaging render. Anything in that image that is not the product (props, background shapes, logos) has to be excluded in the prompt, so keep product references clean.
The trick that makes the composite work: each reference is given exactly one job in the prompt — the character sheet is identity only, the product image is packaging only, the garment image is wardrobe only. Without that, models borrow the background or the pose from the wrong image.
Step 1 — The composite still
Three references go into the chat, the prompt gives each of them one role, and the still comes out. Watch it happen:
References
Create a photorealistic vertical 9:16 image for the opening frame of a skincare UGC video. Reference roles: Use the Ava Marleen character sheet exclusively for identity, the Elora product image exclusively for packaging, and the cream "CHERRY" T-shirt image exclusively for wardrobe. Create one continuous scene. Subject and action: Preserve Ava's reference facial structure, brown eyes, natural brows, nose, lips, warm skin tone, body proportions, and shoulder-length dark hair with its center part and lighter strands. Keep her small gold hoop earrings. She stands comfortably at a bathroom vanity, shoulders relaxed, looking directly into the camera with gently parted lips and a subtle, welcoming expression, captured just before speaking. In her left hand, on screen-right, she holds the closed Elora bottle upright beside her lower cheek, with a small gap separating the bottle from her face. Her fingers support its base and side edges without covering the front artwork. Keep the bottle near the plane of her face, with its label squarely facing the camera and her raised forearm clear of the central shirt graphic. Wardrobe: Dress Ava in the exact cream oversized T-shirt from the garment reference. Preserve its relaxed boxy fit, dropped shoulder seams, loose short sleeves, ribbed crew neckline, and soft cotton texture. Reproduce the centered chest artwork: large burgundy serif "CHERRY" lettering, two illustrated deep-red cherries connected by stems beneath a burgundy ribbon bow, an olive-green leaf, surrounding sparkle motifs, the smaller bow near the lettering, and the handwritten line "— sweet as cherry —" with a small outlined heart below. Match the reference's graphic scale, placement, colors, and illustrated print texture. Let the fabric drape naturally, with gentle folds that keep the lettering and illustration recognizable. Leave the shirt untucked. Product fidelity: Reproduce the reference bottle's proportions, rounded rectangular pink body, dark teal collar, and plum-colored dropper bulb. Preserve the large lowercase dark teal "elora" artwork across the lower front and the white "Retinol Restore" title above "SKIN RENEWAL." Retain the reference placement and typography of "30 ml," "1.01 fl.oz," "PM USE ONLY," "ENCAPSULATED RETINOL," "0.3%," and "SUPPORTS NIGHTLY RENEWAL." Prioritize clear recognition of the brand and product title; preserve the smaller printing at its natural scale. Match all packaging artwork directly to the product reference. Exclude the surrounding blue sculptural forms and the separate symbol at the bottom of that reference image. Environment: A quiet, lived-in home bathroom at night, with warm ivory walls, a pale stone vanity, and a neatly hung off-white towel. An understated ceramic dish sits toward the back of the counter. Keep the setting simple and personal, with clear indoor air and dry surfaces. Depth and composition: A narrow, softly out-of-focus vanity edge anchors the bottom foreground without covering the shirt artwork. Ava and the bottle occupy the sharp middle plane. The tiled wall and towel recede behind her, gently softened but recognizable. Use a waist-up composition that includes the entire printed design down to the small heart, with modest headroom. Place her face slightly left of center and the bottle beside it on screen-right. Keep the bottle at believable handheld scale. Leave clear space around the raised hand for a small subsequent movement. Her face and the product lead the visual hierarchy; the shirt adds a recognizable lifestyle detail. Camera and focus: Eye-level smartphone video-frame appearance, photographed by a phone mounted just beyond the vanity, with a natural 35mm-equivalent field of view. Use a direct camera view with correctly oriented lettering. Keep both eyes and the bottle's front label within the sharp region, and retain enough depth of field for the shirt artwork to remain clear. Allow gentle background separation without artificial portrait-mode cutout edges. Stable framing and natural facial perspective. Lighting: A broad, diffused vanity light just off camera-left and slightly above eye level supplies soft, warm-neutral illumination. Create plausible catchlights in her eyes, gentle highlights across her cheekbones, and soft shadows beneath her chin and toward camera-right. The pale opposite wall provides restrained reflected fill. A faint warm bathroom light illuminates the background. Keep reflections on the bottle broad and controlled so they do not wash out the branding. Preserve visible texture in the cream shirt without clipping its highlights. Mood and realism: Intimate, relaxed evening skincare content with the understated polish of a creator filming at home. Preserve natural pores, subtle under-eye texture, individual eyebrow hairs, fine flyaways, realistic lip texture, and the reference's facial asymmetry. Show believable cotton folds, ribbing at the collar, stitched seams, and a printed design that follows the fabric's surface. Maintain anatomically natural hands, plausible finger pressure, and consistent contact shadows where she holds the bottle. Color treatment: Restrained smartphone processing, gentle contrast, warm ivory surroundings, and accurate skin color. Preserve the shirt's cream, burgundy, and olive hues and the product's pink, dark teal, and plum colors. Keep shadows neutral enough to retain packaging fidelity. Constraints: Preserve the referenced identity, garment design, and packaging while allowing this new pose, expression, setting, and lighting. No tank top, invented freckles or tattoos, extra jewelry, skin smoothing, exaggerated wet skin, distorted hands, duplicated products, floating bottle, mirrored or substituted lettering, cropped shirt artwork, captions, interface elements, reference-sheet borders, or watermarks. Capture one still instant with no motion blur.
Why Nano Banana and not GPT Image 2. Same prompt, same attachments, both models.
Used
Too polished- 1Skin smoothing — pores and under-eye texture are gone; it reads as retouched.
- 2Editorial pose — a three-quarter turn from a model shoot, not a creator facing her phone.
- 3Hotel bathroom — the walls and vanity look staged; the brief was a lived-in home.
- 4Warmer grade — model-shoot lighting and colour instead of flat vanity light and phone processing.
GPT Image 2 gave a beautiful image and the wrong one: smoother skin, an editorial three-quarter pose, a hotel bathroom, warmer grade. Nano Banana kept the bare walls, the flat vanity light and the phone processing. For a UGC ad, imperfection is the brief.
Step 2 — The video
The still becomes the first frame. One reference in, one video prompt, the UGC video out:
References
The clip is image-to-video in Seedance 2.0 from the still. Two rules did the work. The camera stays static — a phone on a mount does not move, and a moving camera is the first thing that reads as AI. And one job per shot: the real video is delivery (talk, bottle parked). Stacking a camera move, blocking and dialogue in one shot is how you get face drift and invented hands.
The optional hook uses the same still with a different job: blocking (talk, then push the bottle to the lens). Generate it only if the cut needs a lead-in.
Static phone camera mounted just beyond the bathroom vanity, eye level, 35mm-equivalent, locked framing with only faint natural micro-shake. Start exactly on the attached frame: the woman in the cream oversized "CHERRY" T-shirt holds the closed pink elora Retinol Restore bottle upright beside her cheek, label facing camera, bathroom lit by one warm vanity light from camera-left. She talks directly to the lens like a creator filming her night routine, relaxed and unscripted: "[SCRIPT — e.g. A little skincare before calling it a night. This is elora Retinol Restore, a nighttime serum. Keeping tonight simple — take a closer look at elora.]" Performance: natural speech rhythm with tiny pauses, eyebrows and eyes doing most of the expression, a small nod early in the line, a quick genuine smile in the middle, a slight head tilt on the product name, and a soft shrug on the last line. She blinks naturally. Her free hand makes one small open-palm gesture around the middle of the line, then relaxes. Her weight shifts slightly once. The bottle hand stays parked beside her cheek the entire time, label square to camera, fingers steady on the base and side edges, never covering the "elora" lettering; the bottle drifts no more than a few centimetres with her breathing. End with her closing her mouth and holding a calm look into the lens, locked frame. Real time, no slow motion, no camera movement, no zoom. Her face stays identical to the start frame throughout, no drift, no morphing. Label lettering stays exactly as in the frame, correctly spelled, undistorted. No second bottle, no extra hands, no new people, no background change, no captions or UI. Ambient: quiet bathroom room tone, natural voice with slight room reflection, no music.
Static phone camera mounted just beyond the bathroom vanity, eye level, 35mm-equivalent, no camera movement, only faint natural micro-shake as if resting on a soft mount. Start exactly on the attached frame: the woman stands at the vanity in the cream oversized "CHERRY" T-shirt, holding the closed pink elora Retinol Restore bottle upright beside her cheek, label facing camera. 0–4 s: She looks straight into the lens and speaks naturally, a small breath before the first word, eyebrows lifting slightly, a half-smile at the end of the line: "[HOOK LINE — e.g. This is elora Retinol Restore, a nighttime serum.]" Her head makes small natural talking movements; she blinks once. The bottle stays beside her face, steady. 4–9 s: On the last word, the trigger: she pushes the bottle straight toward the lens in one smooth, confident arm extension, wrist rotating so the front label ends up square to the camera. The bottle grows until it fills roughly two thirds of the frame; her face falls out of focus behind it. The phone's autofocus hesitates for a fraction of a second, then racks cleanly from her face to the label. Fingers keep their grip on the base and side edges and never cover the "elora" lettering or the "Retinol Restore" title. 9–15 s: Hold. The bottle stays locked in the close-up, label sharp and readable, tiny breathing movement in her arm, warm vanity light catching the pink body with one broad soft highlight. She stays silent. End on this locked frame. Real time, no slow motion, no zoom. Her face stays identical to the start frame throughout, no drift, no morphing. The T-shirt graphic stays unchanged where visible. Bottle label lettering stays exactly as in the frame, correctly spelled, undistorted, never mirrored. No second bottle, no extra hands, no new people, no camera push-in, no background change, no captions or UI. Ambient: quiet bathroom room tone, soft fabric rustle as the arm extends, no music.
Script lines are slots. Write your own; keep them short enough to say in one breath.
Make it yours — the template kit
Fill the ten slots with your character, product, garment and setting; the prompts rewrite themselves. Copy each one into the model it is written for.
Image model. Attach the character sheet, the product image and the garment image, in that order. 9:16.
Create a photorealistic vertical 9:16 image for the opening frame of a UGC video. Reference roles: use the Ava Marleen character sheet exclusively for identity, the elora Retinol Restore product image exclusively for packaging, and the garment image exclusively for wardrobe. Create one continuous scene. Subject and action: preserve Ava Marleen's reference facial structure, brown eyes, natural brows, warm skin tone, shoulder-length dark hair with a center part and lighter strands, small gold hoop earrings, and body proportions. She stands comfortably in quiet home bathroom at night, warm ivory walls, pale stone vanity, off-white towel, shoulders relaxed, looking directly into the camera with gently parted lips and a subtle, welcoming expression, captured just before speaking. In one hand, on screen-right, she holds the closed elora Retinol Restore upright beside her lower cheek, with a small gap between product and face, fingers on the base and side edges, front artwork uncovered, label squarely facing the camera. Wardrobe: dress her in the exact garment from the garment reference: cream oversized T-shirt, boxy fit, dropped shoulders, ribbed crew neck, centered burgundy "CHERRY" artwork with two cherries, a bow, an olive leaf, sparkles and "— sweet as cherry —". Match the reference's graphic scale, placement, colors and print texture; let the fabric drape naturally and keep the artwork fully readable. Product fidelity: reproduce the reference product exactly: rounded rectangular pink dropper bottle, dark teal collar, plum dropper bulb, large lowercase dark teal "elora" on the lower front, white "Retinol Restore" title above "SKIN RENEWAL". Retain the placement and typography of "30 ml", "1.01 fl.oz", "PM USE ONLY", "ENCAPSULATED RETINOL", "0.3%", "SUPPORTS NIGHTLY RENEWAL". Prioritize clear recognition of the brand and product title. Ignore any background, props or symbols in the product reference that are not the product itself. Environment: quiet home bathroom at night, warm ivory walls, pale stone vanity, off-white towel, simple and personal, clear indoor air, dry surfaces. Depth and composition: a narrow, softly out-of-focus counter edge in the bottom foreground; the subject and product on the sharp middle plane; the room receding behind her, softened but recognizable. Waist-up composition that includes the whole garment artwork, modest headroom, face slightly left of center, product beside it on screen-right at believable handheld scale, clear space around the raised hand for a small later movement. Camera and focus: eye-level smartphone video-frame look, phone mounted just beyond the counter, natural 35mm-equivalent field of view, correctly oriented lettering, both eyes and the product label sharp, enough depth of field for the garment artwork, no artificial portrait-mode cutout edges. Lighting: broad diffused vanity light off camera-left, slightly above eye level supplies soft, warm-neutral illumination: plausible catchlights, gentle highlights on the cheekbones, soft shadows under the chin and toward camera-right, restrained reflected fill from the opposite wall, broad controlled reflections on the product that do not wash out the branding. Mood and realism: intimate, relaxed creator content filmed at home. Preserve natural pores, subtle under-eye texture, individual brow hairs, fine flyaways, realistic lip texture and the reference's facial asymmetry; believable fabric folds and seams; anatomically natural hands with consistent contact shadows. Color treatment: restrained smartphone processing, gentle contrast, accurate skin color, the garment's and product's true colors preserved. Constraints: preserve the referenced identity, garment and packaging while allowing this new pose, expression, setting and lighting. No invented freckles, tattoos or jewelry, no skin smoothing, no distorted hands, no duplicated or floating products, no mirrored or substituted lettering, no cropped garment artwork, no captions, UI, reference-sheet borders or watermarks. One still instant, no motion blur.
Checklist before you generate:
- the character sheet has four views;
- the product reference contains only the product;
- the garment is a flat;
- the prompt gives each reference one role;
- the video prompt opens with a static camera and does one thing;
- the script fits in one breath.
Closing
This took an afternoon, and most of that was writing the prompts you just copied. If you want this for your brand — a run of variations across products, creators and hooks — talk to us.
The NOQTA Team