Most first attempts at an AI product ad go the same way. You type "a cinematic ad for my water bottle", generate one long clip, and get something pretty that does not sell anything. The bottle has the wrong logo, the shot never shows what makes the product different, and the clip does not fit the placement you bought.
A better single prompt will not fix this. Ads are cut from short, planned shots, each with one job, and current AI video models produce clips of 4 to 15 seconds, which suits that structure. This guide covers the workflow step by step and ends with a 15-second shot list you can copy.
Step 1: Write a One-Paragraph Brief
Before you open a generator, answer four questions in writing. Keep the answers short.
- Who is the ad for? One audience, described as a person. "Runners who train before work" is useful. "Everyone" is not.
- What is the one message? One benefit, not five. "Keeps water cold for a full day" beats a list of features.
- Where will it run? The placement decides the aspect ratio, the length, and whether people watch with sound.
- What should the viewer do? Visit a page, use a code, or remember the name.
Here is an example brief:
Audience: runners who train early in the morning.
Message: the bottle keeps water ice cold through a long run.
Placement: vertical feed ads, often watched without sound.
Action: tap through to the product page.
Length: 15 seconds.
When a shot looks great but does not serve the message, cut it.
Step 2: Turn the Brief Into a Shot List
A 15-second ad usually holds four or five shots. Each shot has one job: hook, show the product, prove the benefit, show it in real life, and close on the brand.
Write the list before you generate anything. For each shot, note:
- Purpose: what this shot must make the viewer feel or understand.
- Source: does it start from your product photo (image-to-video) or from text alone (text-to-video)?
- Length: most AI models generate clips of a few seconds, so plan shots of 4 to 6 seconds and trim in your editor.
- Camera: one clear move per shot. For a full vocabulary of moves, see our camera movement prompt guide.
Step 3: Hero Shots From Your Product Photo (Image-to-Video)
Any shot where the product must look exactly right should start from a real product photo. Text alone cannot reproduce your logo, your label, or the exact shape of your packaging. A photo can.
Several models on veo4.dev accept a starting image. Veo 4 takes a reference image. Veo 3.1 takes a first frame and a last frame, so you control where the shot starts and where it ends. MiniMax H3 also accepts a first and a last frame. Kling 3.0, Kling v2.5 Turbo Pro, Seedance 2.0, and Wan 2.6 accept an optional first frame, and Hailuo 2.3 works only from a source image. See the image-to-video page for the full flow.
Tips for hero shots:
- Describe motion, not the product. The model can already see the bottle. Tell it what changes: light moves, condensation forms, the camera orbits.
- Keep the product motion small. Big product motion is where labels warp. Move the camera and the light, and let the product stay still.
- Crop the photo first. On Kling 3.0, Kling v2.5 Turbo Pro, and MiniMax H3, the clip takes its aspect ratio from the starting image, so crop your product photo to 9:16 before you upload it for a vertical ad.
- Use a negative prompt on Kling v2.5 Turbo Pro to rule out warped text and extra objects.
The camera slowly orbits the steel water bottle on the wet stone surface.
Morning light sweeps across the brushed metal and catches the logo.
Cold condensation droplets bead and slide down the side of the bottle.
Static product, smooth camera motion, shallow depth of field.
If the label still shifts between takes, our guide to fixing common AI video problems covers warped text and flicker in detail.
Step 4: Lifestyle Shots With Text-to-Video
Lifestyle shots show the product in a life your audience wants. They sell the feeling, and the product is often small in frame or out of focus, which makes them a good fit for text-to-video. Describe who, where, what light, what action, and which camera move.
A young woman jogs along a quiet riverside path at sunrise, breath visible
in the cold air. She slows down, lifts a steel water bottle and drinks.
Golden backlight, light mist over the water. Handheld tracking shot at
shoulder height, natural and slightly shaky.
If the same person appears in two or more shots, you need a plan to keep them consistent. Our guide on consistent characters across AI video clips explains how.
Step 5: Voiceover, Sound, and Music
Many feed ads play without sound, so the picture must carry the message on its own. Add on-screen text in your editor for the key benefit. Then treat sound as a bonus that makes the ad better for people who do listen.
You have two options:
- Native audio. Some models generate sound in the same pass as the picture. Seedance 2.0 has a native audio switch, and Kling 3.0 has a native sound switch. Use native audio for ambience and effects: footsteps on gravel, a cap twisting open, birds at dawn.
- Separate voiceover and music. For a scripted line of copy or a brand track, record or generate the voiceover separately and lay it over the cut in your editor. This keeps the voice the same across every shot.
Describe sound in the prompt with the same care as the picture:
Sound: the soft click of the cap unscrewing, a short splash of water,
quiet morning birdsong in the background. No music, no dialogue.
Step 6: Match the Aspect Ratio to the Placement
Generate each shot in the aspect ratio of its placement. Cropping a wide clip to vertical later throws away most of the frame and often cuts off the product.
| Placement | Aspect ratio | Notes |
|---|---|---|
| TikTok, Reels, Shorts, Stories | 9:16 | Keep the product in the middle; the top and bottom hold app buttons |
| Feed posts on most platforms | 1:1 or 3:4 | Square is safe everywhere; 3:4 takes more screen height |
| YouTube in-stream, websites | 16:9 | The classic wide format |
| Banner-style or cinematic headers | 21:9 | Available on some models, such as Seedance 2.0 and MiniMax H3 |
Every text-to-video model on the site with an aspect ratio setting offers 16:9 and 9:16. Seedance 2.0, MiniMax H3, and Happy Horse 1.0 also offer 1:1, 4:3, and 3:4. Veo 4 and Veo 3.1 offer only 16:9 and 9:16. If you need vertical and wide versions of the same ad, generate each version natively rather than cropping one. For more on vertical framing, read making vertical AI videos for TikTok, Reels, and Shorts.
Step 7: Iterate Cheaply, Then Render the Finals
On most models we offer, the credit cost of a clip grows with its length, and on many it also grows with resolution. Native audio raises the cost on Seedance 2.0 and on the standard and 1080p modes of Kling 3.0. Use that to your advantage:
- Draft short. Use the shortest duration the model offers, such as 4 seconds on Veo 4 and Veo 3.1 or 5 seconds on Kling and Seedance 2.0.
- Draft at a lower resolution. On models whose cost depends on resolution, such as Seedance 2.0, Kling 3.0, and Wan 2.6, test the prompt at 480p or 720p. On Veo 4 and Veo 3.1 the cost depends only on duration, so resolution does not change the price.
- Change one thing per take, so you know which change helped. Keep the starting image, the duration, and the aspect ratio the same between takes.
- Render the finals last. Once a prompt works, run it at full length and 1080p, or higher on models such as Kling 3.0, which offers 4K.
Check the current credit cost of each option on the pricing page before you plan a batch.
A Sample 15-Second Ad: Shot List With Prompts
Here is a complete plan for the water bottle brief above, in 9:16 for vertical feeds.
| Shot | Time | Job | Source |
|---|---|---|---|
| 1 | 0 to 3 s | Hook | Text-to-video |
| 2 | 3 to 7 s | Hero product | Image-to-video |
| 3 | 7 to 10 s | Proof of benefit | Image-to-video |
| 4 | 10 to 13 s | Lifestyle | Text-to-video |
| 5 | 13 to 15 s | Brand close | Image-to-video |
Shot 1, the hook:
Extreme close-up of an ice cube dropping into clear water inside a steel
bottle, in slow motion. Water splashes up toward the lens. Cool blue light,
dark background. Static camera. Sound: a crisp ice clink and splash.
Shot 2, the hero shot (start from your product photo):
The camera slowly pushes in on the steel water bottle standing on a wet
stone. Soft sunrise light slides across the brushed metal and the logo.
The product stays still. Shallow depth of field, calm and premium mood.
Shot 3, proof of the benefit (start from your product photo):
Cold condensation forms on the outside of the bottle and a single drop
runs down past the logo. Bright midday sun behind it, heat shimmer in the
background. Static close-up shot. Sound: faint summer insects.
Shot 4, lifestyle:
A runner stops at the top of a hill at sunrise, unscrews a steel bottle
and drinks, then smiles and looks out over the city. Warm backlight, light
breeze in her hair. Slow handheld orbit around her at shoulder height.
Sound: heavy breathing, wind, distant birds.
Shot 5, the brand close (start from a clean packshot):
The bottle stands centered on a plain light background. The camera holds
still while soft light slowly brightens from left to right. Leave clear
empty space above the bottle for a logo and a tagline.
Add the benefit as on-screen text over shots 2 and 3, the call to action over shot 5, and one voiceover line across the whole ad.
FAQ
Which model should I use for product ads?
For shots where the product must look exact, use a model that accepts a starting image, such as Veo 4 or Veo 3.1. Use Veo 3.1 when you also want to fix the final frame. For lifestyle shots, any text-to-video model works.
Can AI reproduce my exact logo and label?
Start from a real product photo and keep the product motion small. Text-only prompts cannot reliably reproduce brand marks. For a perfect logo, add it in your editor on the final frame.
Should I use native audio or a separate voiceover?
Use native audio for the sounds inside each shot. Use a separate voiceover and music bed across the whole ad, so the voice and the track stay the same from shot to shot.
How do I make a vertical and a wide version?
Generate each version in its own aspect ratio, 9:16 and 16:9, from the same prompts. Native framing keeps the product in the frame; cropping often cuts it off.
Start Your First Product Ad
Write the brief, plan five shots, and draft them short. Then open the AI video generator and start with your hero shot, or go straight to Veo 4 to animate your product photo. When the drafts look right, render the finals and cut the ad together.