Skincare and beauty prompts with the same face and the same bottle

What makes beauty prompts work

Keep two reference sets: the face, photographed bare or with light makeup in soft even light, and the product, with a sharp close-up of the label. Heavy retouching in face references carries into the video as plastic skin.

Ask for natural skin. Words like "visible skin texture, soft diffused light, no heavy retouching" push the model away from the airbrushed look that makes beauty clips feel fake.

Choose motions the model can do. Fingers patting cream onto a cheek works well. Mascara, eyeliner and anything close to the eye often fails, because tiny tools near small features are hard to keep stable.

Length, shape and sound are in brackets before each prompt; they are settings, not prompt text. One-subject prompts use a role, such as "the serum bottle". Prompts with several subjects name them: Lena, LUMEN, HALO and MIST are example names, so write your own saved names there.

Product and texture shots

Prompt (5 s, 9:16): The serum bottle stands on a wet white stone surface. A glass dropper rises out of it and releases a single clear golden drop that lands on the stone and spreads slightly. Macro lens, soft top light with a bright highlight on the glass, pale pink background. Slow motion.

Why it works: one drop is a simple, satisfying motion that shows the texture. If the dropper changes shape, include a photo of the dropper out of the bottle.

Prompt (6 s, 9:16): Extreme close-up of a fingertip swirling a scoop of cream from the jar, lifting it to show a soft peak, then smoothing it flat. Diffused daylight from a window, shallow depth of field.

Why it works: a swirl and a peak show thickness, which is what a still photo cannot. Thin lotions read better as a slow drip from the fingertip instead.

Application on a consistent face

Prompt (6 s, 9:16): The woman, hair pulled back with a soft headband, gently pats moisturizer into her left cheek with two fingers, then turns her face slightly to the window. Close-up, static camera at eye level. Soft morning daylight, visible skin texture, calm expression.

Why it works: patting is slow and repetitive, so hands stay stable. The slight turn shows the glow on the skin without a large head movement.

Prompt (6 s, 9:16): The man stands at a bathroom mirror and splashes water on his face with both hands, then looks up at his reflection, water on his skin. The camera is behind his shoulder, focused on the reflection. Bright, clean white light, tiled wall.

Why it works: a mirror shot doubles the face, so the reference set must include a clear front view. If the reflection disagrees with the person, film from the side instead.

Prompt (5 s, 9:16): Close-up of Lena applying HALO lip balm in one slow stroke across her lower lip, then pressing her lips together and smiling slightly. Soft frontal light, neutral background.

Why it works: one stroke and one press are easy to read. Avoid asking for a full lipstick application in one clip; outlines tend to drift.

Routine and lifestyle scenes

Prompt (7 s, 9:16): Lena sits at a small vanity in a bright bedroom, LUMEN and HALO lined up in front of her. She picks up LUMEN, reads the label for a moment and smiles at the camera. Medium shot, warm morning light, plants in the background.

Why it works: lining products up keeps them separate and in view. That is three subjects, the most one shot can hold in RefVid, so give each product its own clear reference photos.

Prompt (8 s, 16:9): Overhead shot of a marble bathroom shelf with MIST, LUMEN and HALO arranged in a row. A hand reaches in and lifts each one in turn, then sets it back in place. Soft shadowless light.

Why it works: overhead flat lays are clear and brand friendly. If items swap positions, reduce the scene to two products.

Claims, labels and model notes

Avoid before-and-after transformations and medical results in AI-generated beauty ads. Ad platforms often restrict them, and a generated result proves nothing about the product. Show texture, routine and feel instead.

When face and products share a shot, name each one. RefVid stores the face and each product as separate saved subjects, each with 1 to 5 photos, and up to 3 can share one shot.

On Veo 3.1, as of October 2026, according to Google's Veo docs, a shot takes up to three reference images in total, so with one face and one product, give the face two slots.