An AI video ad generator can turn a product photo into short moving footage. To make a usable ad, start with your own clear photo, generate two simple product shots, check that the item still matches what you sell, then add the exact logo, price, call to action, and voiceover in an external editor. Review the exported file in its intended placement.
The workflow below builds a 12-second vertical ad from one photo of a closed jar candle: two generated shots and an original-photo end card. It is a copy-ready shot-plan and prompt example, rather than a tested product video. Replace the candle details with your actual item.
1. Choose a photo that can carry the ad
Use a merchant-owned or properly licensed photograph showing the complete product. For this example, choose an unlit candle in an amber-glass jar with a black lid and a front-facing label. If your product differs, change every description to match it.
Check the source at full size: readable label, complete lid and base, clear glass color. Avoid store-interface screenshots or promotional badges. Keep the untouched original for comparison.
Prepare a vertical composition with space for text above or beside the product. Extend a plain background externally if needed; keep important details away from the edges.
Write down three things before generating:
- Product identity: silhouette, proportions, color, lid, label position, and exact visible wording
- Approved message: one product fact supported by the listing or your own evidence
- Offer: current price, currency, conditions, destination page, and one next step
A photograph documents appearance. It does not establish burn time, fragrance strength, safety, or material performance. Use verified facts for those claims and real demonstration footage when the ad needs to prove an action.
2. Plan two moving shots and one end card
Keep the first ad small enough to review. A product reveal, a quiet detail beat, and a readable closing card give each part a clear job.
| Final edit | Visual | On-screen copy example | What must stay accurate |
|---|---|---|---|
| 0–4 seconds | Shot A: full jar, very small push-in | Meet [product name] | Complete silhouette, lid, label, color |
| 4–8 seconds | Shot B: same front view, fixed camera, gentle background movement | Amber glass. Black lid. | Use this wording only if it matches your item |
| 8–12 seconds | Original photo, approved logo, optional current price | Shop [product name] | Exact offer, product name, and destination |
These times describe the external edit. Generate source clips using available durations, then trim them to the timeline. A time instruction in a prompt does not guarantee an exact cut.
The second shot can also show another verified visible detail, such as fabric texture on a bag. With only one photo, avoid requesting a full rotation, an opened container, or the unseen underside. Those views need real supporting material.
3. Set up the image-led generation
Open Seedance 2.5 Image to Video, upload the prepared photo, and check the selected model and input mode. The public workspace may open with Seedance 2.5 Lite selected; explicitly choose the model you intend to use.
Confirm the available duration, vertical 9:16 setting, output option, and displayed credit quote before submitting. If the selected workflow offers generated audio, turn it off when you plan to build the soundtrack in your editor.
Use the photo as the product appearance reference. If the interface assigns an asset tag, replace “the uploaded product photo” in the prompts with that displayed tag. Recheck it after changing uploads.
Save the photo, prompt, settings, and quote together. Reuse the same approved source for both shots.
4. Copy these product-shot prompts
Shot A: introduce the whole item
Use the uploaded product photo as the reference for the closed candle jar,
its starting composition, amber glass, black lid, and front label.
Create one four-second product shot. The jar remains stationary and unlit.
Make one very small, slow camera push-in, then settle into a steady hold.
Keep the entire lid, jar outline, label, and base inside the frame.
Preserve the visible proportions, glass color, lid shape, label position,
and printed wording from the source. Keep the original light direction.
No rotation, lid opening, flame, hands, extra products, new markings,
new text, scene change, or additional cuts. Leave the planned text area clear.
Shot B: give the detail room to breathe
Use the same uploaded product photo for the exact candle jar and front view.
Create one four-second product shot. The closed, unlit jar stays completely still. Use a fixed camera and keep
its complete outline visible. Preserve the lid, glass color, label edges,
logo, and visible printed wording from the source.
A faint, soft light pattern moves slowly across the empty background only.
Keep the product's exposure and color stable. End on a steady composition.
No zoom, orbit, turning jar, flame, hands, extra objects, new text, or cuts.
These instructions define what to request and inspect. They cannot guarantee exact typography. If background light distracts from the label, remove that instruction and keep Shot B still.
5. Check the logo and packaging before editing
Compare each downloaded clip with the original photo at the opening, middle, closest view, and ending. Then scrub between those checkpoints. A label can look correct in a paused frame and briefly change during movement.
Read the visible product name letter by letter. Check the logo shape, line breaks, label borders, jar height-to-width ratio, lid thickness, and number of visible parts. Check full-size and phone-size playback.
Reject a changed ingredient, quantity, warning, certification, or brand name. Adding a clean logo in the corner does not repair incorrect printing on the product itself.
| Problem | First useful change | When to use the original photo instead |
|---|---|---|
| Label letters wobble during the push-in | Reduce the push-in or request a fixed camera | Essential wording still changes |
| Lid or jar shape changes | Remove rotation and simplify movement | The product no longer matches the listing |
| Product leaves the frame | Add source-image margin and reduce camera travel | The required edge is missing from the source |
| Unrequested flame or extra objects appear | Explicitly keep the jar closed and unlit | The unwanted action cannot be cleanly trimmed |
Change one cause at a time and record the first problem's timestamp. Keep a passing shot when only the other shot needs replacing. If identity remains unreliable, use the original photograph with a restrained editor move for that segment.
6. Add exact copy, price, and voiceover externally
Place the two accepted clips on the timeline and finish with the original photo. Add your approved logo as a separate asset. Type the product name, current price, and CTA in the editor so a copy change does not require another generation.
Keep overlays short, high-contrast, and clear of the label. For TikTok placements, use the applicable safe-zone template: the usable area changes with captions and other interface elements. Preview the actual layout rather than assuming every vertical placement has identical margins.
A simple narration template is:
“Meet [product name]. [One verified product fact]. Shop [brand].”
Record it yourself or use a voice you have permission to use. Check product pronunciation and every number. Caption the important words, keep music below the voice, and watch the ad muted to confirm the product and CTA remain understandable.
Use a price only when the landing page offers the same item and conditions. Save an offer-free version for reuse.
7. Budget for attempts, then verify the export
This website has a freemium starting path with free credits. Check your account's current balance and eligible options; this workflow does not assume a fixed allowance or that it covers both shots. Review export access and commercial-use terms before planning a paid campaign.
Write down each shot's live quote. Your credit budget is the quote for Shot A multiplied by its planned attempts, plus the equivalent calculation for Shot B. The photo end card needs no generation.
For a small first test, budget the two initial shots and a limited replacement allowance. When both shots have the same quote, two initial submissions plus one replacement require three times that quote. This is a planning scenario, not a success-rate estimate. The Seedance 2.5 cost-per-video guide explains how discarded attempts affect usable-output cost.
Stop repeated submissions when the same essential detail keeps failing. Improve the source or use the photo-based edit. Compare the live options on Seedance 2.5 pricing before buying more capacity.
Export a vertical MP4 for the chosen placement and inspect the actual dimensions, duration, audio, and playback. A larger export setting does not restore missing detail; the Seedance 2.5 resolution guide covers that check. Confirm the offer and CTA in the final file, then preview it on a phone with the platform's interface overlays.
FAQ
Can one product photo make a complete ad?
It can support a short appearance-focused ad using reviewed footage and an original-photo end card. Add real evidence when the message depends on product performance or a demonstration.
Will an AI video ad generator preserve my logo exactly?
Do not assume it will. Inspect the packaging throughout the clip, and add exact brand assets externally. Use the original photo when essential printed details cannot be preserved.
Can I make the first ad for free?
Check the available credits, selected model, displayed cost, download access, and intended-use terms. A free starting balance does not establish how many usable shots you will receive.
Start with one product and the two-shot plan in Seedance 2.5. Finish the message in your editor and check the final export.
Sources
- TikTok Auction In-Feed Ads: placement specifications and variable safe zones
- TikTok Creative Best Practices: recommended vertical framing, readable overlays, and a clear CTA
- Google Ads Misrepresentation Policy: misleading product claims, manipulated media, and unavailable offers
