Food and drink video prompts with real texture and motion
Food and drink clips succeed or fail on texture and motion: a pour that behaves like liquid, steam that rises at the right speed, a glaze that catches light. Reference images keep your can, bottle or signature dish looking like yours. The prompts below cover pours, plated dishes and packaging scenes, with notes on the physics the models still get wrong.
What makes food prompts work
Use references for anything branded or signature: the can, the bottle, the box, or a dish your restaurant is known for. Generic ingredients like ice or herbs do not need references; the model knows them.
Light it like a food photographer. Drinks look best backlit so the light passes through the liquid; plated food looks best with soft side light that shows texture. Say which you want.
Give liquids one clear path and a speed. "A thin steady stream poured from six inches above the glass" produces a better pour than "pouring". Slow motion helps, but too much of it makes liquid look like gel.
Each prompt below has its length, shape and sound in brackets; those are settings, not prompt text. The prompts refer to the saved product by its role, such as "the can", and RefVid attaches its photos.
Pours and liquids
Prompt (6 s, 9:16, sound on): The can stands beside a tall clear glass filled with ice on a dark wooden bar. A hand lifts the can and pours a steady amber stream into the glass, foam rising to just below the rim. Backlight from behind the glass makes the drink glow; condensation beads on the can. Static close-up camera. Sound: the hiss of the can opening, fizz in the glass.
Why it works: backlight shows the color and the bubbles. Keep the can label toward the camera as it tilts; if it smears, start the clip already pouring.
Prompt (7 s, 9:16): Overhead shot of the cup on a light stone counter. Milk is poured from a small steel jug into dark espresso, slowly forming a simple heart on the surface. Soft daylight from a window above, gentle steam.
Why it works: the overhead angle turns the pour into a clear, readable shape. A simple heart is far more reliable than a detailed pattern. Here the saved product is your café's branded cup.
Prompt (5 s, 16:9): Close-up of the pancake stack on a blue plate. Thick maple syrup is drizzled from above in a slow ribbon, pooling at the top and running down the sides. Warm side light from the right, shallow depth of field, kitchen blurred behind. Slow motion.
Why it works: syrup moves slowly, so slow motion looks natural here. For thin liquids like water or juice, ask for normal speed.
Plated dishes and close-ups
Prompt (8 s, 16:9): The ramen bowl sits on a dark wooden table in a quiet restaurant at night. Steam rises steadily from the broth while the camera pushes in slowly from a three-quarter angle to a close-up of the egg and noodles. Warm overhead pendant light, deep shadows around the bowl.
Why it works: steam and a slow push-in add life without moving the food itself. Photograph the bowl from the same three-quarter angle for the references.
Prompt (5 s, 16:9): Chopsticks lift a few noodles from the ramen bowl, hold them up toward the camera for a beat, and lower them back. Side light catches the shine of the broth on the noodles. Static macro camera.
Why it works: a lift and return is one motion with a clean ending. If the chopsticks bend or multiply, frame tighter so only the tips show.
Packaging and table scenes
Prompt (6 s, 16:9): The snack bag lies on a checked picnic blanket in dappled shade. A hand reaches in from the left, tears the top open and tips a few chips onto the blanket. Late afternoon sun through leaves, gentle movement of light on the bag. Static camera, slightly high angle.
Why it works: dappled light makes a still product scene feel alive. Crinkled packaging can lose its logo, so include a flat front photo of the bag.
Prompt (8 s, 16:9): The wine bottle rests in a silver ice bucket on a white tablecloth at an outdoor dinner. The camera slowly arcs from the side to the front label while candlelight flickers on the glass. Blue dusk sky, string lights out of focus behind.
Why it works: the arc ends on the front label, the angle the references show best. If ice clips through the bottle, reduce the ice to a few cubes.
Sound and saved dishes in RefVid
Sound sells food: a can hissing, oil sizzling, a spoon on porcelain. In RefVid, turn sound on and describe the sounds in the prompt; Kling 3.0 Omni generates them with the picture. Leave sound off for silent social cuts.
When a scene has a branded product and a second item with its own references, such as a bottle and a branded glass, save each as its own product and write both saved names in the prompt. Up to 3 subjects can share one shot.
RefVid keeps each product or signature dish as a saved subject with 1 to 5 photos, so a menu of ten dishes is ten subjects reused across every seasonal campaign. Clips run 3 to 15 seconds; a full pour fits in 6.
Frequently asked questions
- Why do pours look fake in AI video?
- Usually too much slow motion or no defined stream. Describe the stream thickness and height, use normal speed for thin liquids, and backlight the glass.
- Do I need references for generic food ingredients?
- No. Use references for branded packaging and signature dishes. Generic items like fruit, ice or bread are well known to the models.
- Can AI video show my restaurant's real dish accurately?
- It can keep the look close with good references from several angles. It will not reproduce exact garnish placement every time, so check each clip before using it on a menu.