Guides / Ads

A drink ad in fifteen seconds, from stills you generate

A food or drink ad is three or four shots that make the product look cold, fresh and worth reaching for. You do not need a set. You need one good still, three animated shots and a cut.

  • 10 min read
  • Presets: Hands-on product ad, Commercial
The stillThe pourThe ad

What you make, in order. Tap one to jump to its step.

To make a food or drink video ad with AI, generate one styled still of the product, animate three shots from it (a hero, a pour or motion shot, a macro), then cut them to a sound bed in Studio. That is the whole method. Everything on this page was made that way for a fictional sparkling tonic with no brand, and every clip is the real output of the run described next to it.

The product is invented on purpose: a slim bottle, a black cap and an orange band with a sun mark and no words. Use your own product photo where we used a generated still; the steps are the same.

Before you start

  • One clear photo of the product, or a written description if it does not exist yet. Every shot in the ad is made from one styled still, so get that still right first.
  • Three cues you can name. For a cold drink: condensation running, bubbles rising, a pour over ice. For hot food: steam, a glaze catching light, a cut that shows the inside. The model only renders what the prompt names.
  • Where the ad will run. Every shot below is 16:9. Make it 9:16 from the start for Stories and Reels; a 16:9 clip does not turn into a good 9:16 one afterwards (see the limits below).
  • No words in the picture. Keep lettering off the label and out of the prompt. Add any text as a layer in Studio, where it is typed, not drawn.

Make the styled still

Open the image console and describe the product, the surface it stands on and the light, in that order. If you have a product photo, add it with Add reference image and describe it the same way, so the model keeps its shape and colours. This still is the ad's look: every later shot inherits its gels, haze and contrast.

The still (GPT Image 2.5 Flare, 16:9)
Premium drink advert still, macro product photography. A tall slim clear glass bottle of sparkling yuzu tonic, pale gold liquid with fine bubbles rising, heavy cold condensation beading and running down the glass. Plain matte black cap. The label is only a single band of electric orange with one small abstract sun shape, no lettering at all. The bottle stands on wet black slate. Magenta gel from the left, teal gel from the right, a hard white rim light tracing the bottle edge, thin haze in the air, deep black background, strong contrast, shallow depth of field, 100mm macro lens. No text, no letters, no words, no logos, no brand marks, no watermark, no readable signage.
A slim clear bottle of pale gold sparkling tonic with a black cap and an orange band bearing a small sun mark, covered in condensation, standing on wet black slate with magenta haze on the left and teal haze on the right
Made with GPT Image 2.5 Flare. One run, first result. The label carries a shape, not a word, so nothing can be misspelled.

Tip:Make one more still per shot, from this one. The pour and the macro below each started as a new still made with this one attached as the reference. That keeps the bottle, the band and the light the same across shots without a product library.

Animate three shots from stills

Open the video console, choose Animate an image, press Add start frame and then Upload image (or Pick an asset for a still already in your Library). Pick the model with the Choose model button; we used Kling v3 (Kling 3.0 in the menu) for all three shots because it keeps a pour, the ice and a standing bottle believable. Set the length on the duration chip. The Generate button shows the live price before you press it.

The video console on a desktop: the pour still attached as the start frame, Kling 3.0 chosen, 4 s, 16:9, 720p, and the Generate button showing the live price
The video console on a desktop. The pour still is the start frame; the price on the button is read live.
The same form on a phone: Kling 3.0, the start frame, duration, aspect ratio, quality and the Generate button with its price
The same form on a phone.
The pour still: the bottle tilted from the upper left pouring pale gold tonic into a glass of ice, droplets frozen in the air, magenta and teal haze
The pour still, made with the bottle still as the reference. It is the start frame of the pour shot.
The macro still: the bottle's shoulder covered in condensation, one large drop running over the top edge of the orange band
The macro still, the same way. It is the start frame of the macro shot.

Write the prompt as motion only. The still already holds the look, so say what moves, what stays and what the sound is. Say what must not move, too: the bottle, the label, the camera.

Hero (Kling v3, 6 s, from the bottle still)
Premium drink advert, one continuous shot from a locked-off camera with only a very slow, gentle push-in that stops early: the whole bottle and its orange label stay fully in frame for the entire shot. Cold condensation beads swell and run down the glass, fine bubbles rise through the pale gold sparkling tonic, the magenta and teal haze drifts slowly behind it, the hard white rim light stays on the bottle edge, the wet slate glistens. The bottle itself does not move or turn. Photoreal, strong contrast, shallow depth of field. No text, no letters, no words, no logos, no brand marks, no watermark. Sound: a soft fizz of bubbles and a low cinematic hum, no music, no voice.
Pour (Kling v3, 4 s, from the pour still)
Premium drink advert, one continuous shot, locked-off camera. The stream of pale gold sparkling tonic keeps pouring from the tilted bottle into the glass of ice, splashing and fizzing at the surface, droplets flying and catching the magenta and teal light, bubbles rising, the haze drifting slowly. The bottle and the glass stay exactly where they are and keep their shape; the orange band and its sun mark stay on the bottle. Photoreal, strong contrast, macro lens. No text, no letters, no logos, no watermark. Sound: pouring liquid and fizz, no music, no voice.
Macro (Kling v3, 4 s, from the macro still)
Premium drink advert, one continuous extreme macro shot, locked-off camera. The large drop of condensation slides slowly down the cold glass and over the top edge of the orange band, smaller beads tremble and merge, fine bubbles rise inside the pale gold tonic, the magenta light on the left edge and the teal on the right flicker very slightly. Nothing else moves; the bottle stays exactly where it is. Photoreal, razor thin depth of field. No text, no letters, no logos, no watermark. Sound: a faint fizz, no music, no voice.
The hero: the tonic bottle on wet slate between magenta and teal haze, the whole bottle and its orange band in frame, the cap closed
The hero. Kling v3, 6 seconds, from the bottle still. The whole bottle and its band stay in frame for all six seconds, the haze drifts behind it, and the cap stays closed.
Pale gold tonic pouring from the tilted bottle into a glass of ice, splashing, then filling the glass to a foam line under magenta and teal light
The pour. Kling v3, 4 seconds, from the pour still. The stream splashes over the ice, then fills the glass to a foam line; the bottle and the band keep their shape.
Extreme close-up of a condensation drop sliding down the bottle over the edge of the orange band
The macro. Kling v3, 4 seconds, from the macro still. The drop slides down over the band and the smaller beads tremble.

Appetite cues: what worked and what did not

CueResult on Kling v3What to do
Condensation running down the glassBeads on the glass in every shot; in the macro the drop slides over the bandName it in every shot that shows the bottle
Bubbles rising inside the drinkVisible through the glass in the hero and the macroAsk for fine bubbles rising, not just sparkling
A pour over ice with splashWorked from a pour still: a splash over the ice, then the glass fills to a foam lineMake a still of the pour first, then animate it
A slow push-in on the heroHeld: the whole bottle stayed in frame. Veo 3.1 Fast walked off the label after 2.5 secondsSay the push-in stops early and the whole bottle stays in frame
A closed bottle standing stillHeld: nothing entered the bottle. Veo 3.1 Fast poured liquid in through the capSay the cap stays closed and nothing enters the frame
Five cues, three shots on Kling v3 at 720p, from the same stills and prompts as the Veo 3.1 Fast takes below.

Veo 3.1 Fast, take 1: walks off the label

The first hero take: a push-in that keeps going until only the bottle's shoulder fills the frame

Veo 3.1 Fast, take 2: a pour through the cap

The second hero take: the bottle stays in frame, but for two seconds liquid pours in through the closed cap
Veo 3.1 Fast, from the same bottle still and prompt, for comparison. The push-in kept going until only the shoulder filled the frame, and the next take poured liquid in through the closed cap for two seconds. Kling v3 held both, which is why every shot above is on it.

Cut it in Studio over a sound bed

Make the music first: open the audio page, choose Music, describe the bed and generate. We used Lyria 3.5. Then open the hero clip from your Library with Open in Studio, switch the media panel to Library, and for each clip set Mark In and Mark Out in the source preview before pressing + to add it at the playhead: the hero from 2.6 s to the end, the pour, the macro, the first 2.5 s of the hero again, then the bottle still as an end card. Add the music the same way from the start. Press Text to add a text clip over the end card and type the line; it is typed, so it cannot be misspelled. Export video renders the whole timeline as an MP4.

The sound bed (Lyria 3.5)
A premium drink advert sound bed, about twenty seconds: a minimal deep electronic pulse, warm sub bass, a crisp shaker groove, a bright synth swell that builds through the middle and resolves with a clean stop at the end. Modern, cool, confident. Instrumental only, no vocals.
The finished 20 second ad: the hero, the pour, the macro, the bottle again, then the still with the typed line Cold. Sharp. Sunlit.
The finished ad. Four clips and one still, cut in Studio, the line typed on the end card, exported as MP4 at 1920 by 1080.
Studio on a desktop: the cut across five tracks with the music underneath
Studio on a desktop.
Studio on a phone: the same timeline, the program monitor on top
The same timeline on a phone.

Which tool for which job

JobUseWhy
A hands-on demo with no presenter, fastHands-on product ad presetOne form, phone-footage look, sound included
A story ad with people and an end cardCommercial presetPlans the beats for you; costs more because it renders several shots
The look of the adImage console, GPT Image 2.5 FlareOne still sets the light, the gels and the product for every shot
Each shotVideo console, Animate an image, Kling v3Keeps a pour and a standing bottle believable
The musicAudio page, Music, Lyria 3.5One prompt, one bed, then trim it in Studio
The cut and the exportStudioTrim, order, audio track, export as MP4

Where you still regenerate

  • The camera does what it wants. On Veo 3.1 Fast two of four takes moved the camera more than asked; on Kling v3 it held. Lock it off in the prompt either way.
  • A new size is a new render. Reframe works on stills, not clips. Decide 16:9 or 9:16 before you animate anything.
  • A label with words will not survive. Keep lettering off the product in the still and add it as text in Studio.
  • Sound is per shot. We made these shots without sound, so the music bed carries the ad. Kling v3 renders its own sound in the console; mute a clip's track in Studio if it fights the beat.

What it costs

This ad took three stills, three clips on Kling v3 (a hero, a pour, a macro) and one music run. Each price is shown on the Generate button before you press it; a render that fails is refunded. Cutting and exporting in Studio added nothing to the bill on our run.

  • Hands-on product ad

    39 credits per 5s at 720p (23 at 480p for drafts, 70 at 1080p)

    Takes 1 to 3 min

  • GPT Image 2.5 FlareEvery plan

    Per image

    Native 8 credits · 2K 21 credits · 4K 58 credits

  • Kling v3Pro and up

    Per 5s clip (audio on by default)

    720p 35 credits · 1080p 47 credits · 4K 117 credits

  • Lyria 3.5Every plan

    Per generation

    Standard 5 credits

Read live from the catalog when this page loads. Every price is shown before you generate. See every rate.

Checklist before you post

  • The product keeps its shape, colour and label mark in every shot.
  • Every shot names one appetite cue and shows it.
  • The camera stays on the product; nothing walks out of frame.
  • Any words are typed in Studio, not drawn by the model.
  • The cut is the length the placement needs, and 9:16 if it runs in Stories.

Try it yourself

Download the bottle still, open the video console, choose Animate an image, upload it as the start frame, pick Kling v3 (Kling 3.0 in the menu), paste the hero prompt above and set 6 seconds. The hero on this page came from that still and that prompt.

The inputs

  • The bottle still (start frame for the hero)JPEG photo · 2560x1440 · 518 KBDownload
  • The pour still (start frame for the pour)JPEG photo · 2560x1440 · 733 KBDownload
Both stills are generated, so you can use them freely to practise.

Questions and answers

Can I do this with my real product?
Yes. Add your product photo as the reference when you make the styled still, and describe its shape, colours and label. Check the still against the real thing before you animate it, because every shot inherits it.
How many shots does a 15 second ad need?
Four or five. Ours is the tail of one hero take, the pour, the macro, the start of the other hero take and the still as an end card, each trimmed to 2.5 to 5 seconds in Studio, 20 seconds in all.
Which model should I animate with?
Kling v3 for every shot: it kept the pour, the ice and the closed bottle believable. Seedance 2.5 takes the same start frame and lets you add reference images if you need the product held from more angles. Veo 3.1 Fast poured through the closed cap on our run.
Can I make the same ad vertical?
Not from these clips. Reframe works on stills only, so make the still and the shots at 9:16 from the start for a vertical placement.
Can the label carry my brand name?
Keep words off the label in the still and add them as a text layer in Studio, where they are typed. A word drawn by the model can come back misspelled and will change between shots.
Does the same method work for hot food?
Yes, with different cues: steam rising, a glaze catching the light, a cut that shows the inside. Name each one in the prompt; we have only run the cold drink on this page.

Make your first shot from a still

Open the video console

Ads

From one product photo to a video ad and a poster set

One clean product photo in; a six second ad and three poster sizes out. The image console restyles the photo, the video console animates it, and the posters keep the same gloves and the same words.

· 8 min read