Dropshipping has a creative problem that other e-commerce models do not: you need advertising for a product that is not in your hands, that you may drop next week, and that twelve other stores are selling with the same supplier photos.
The constraint shapes everything. Here is how to work inside it without doing the things that get accounts shut down.
The thing that actually kills dropshipping ad accounts
Before anything else, because it is the expensive mistake.
Using the supplier's video. The clip on the AliExpress or CJ listing is not licensed to you. Every other store selling that product is also running it. Three consequences, in increasing severity:
- Ad platforms detect duplicate creative across advertisers and suppress it.
- Viewers have seen it, so it does not perform.
- The copyright holder — sometimes the original brand whose product was cloned — issues a complaint, and account restrictions follow.
The same applies to screen-recording a competitor's TikTok, pulling clips from YouTube reviews, and reusing influencer content you did not commission.
This is not a grey area and it is not a risk worth the saved afternoon.
What you can legitimately work from
Supplier still images — usually fine to use, but check the terms; many suppliers explicitly permit resale use of listing images. Stills are a weaker signal of originality than video, so transform them rather than posting them unchanged.
Your own sample. Order one. Yes, it costs two weeks and $15. It is also the only route to genuinely original footage, and for any product you are seriously scaling it pays back immediately.
Licensed stock for context and B-roll — never as the whole ad.
Generated video from the supplier stills. Legitimate, original output, fast enough to test many products. The limits are real and covered below.
Creator content you commissioned. Most authentic, slowest, and the right answer once a product proves out.
The testing phase versus the scaling phase
The central insight for dropshipping specifically: these two phases want completely different creative, and most people use one approach for both.
Testing — finding a product that works
You are testing fifteen products this month. Thirteen will fail. Spending a week on creative for each is irrational.
What you need: fast, cheap, good enough to read a signal. Three variants per product, produced in an hour, not a week.
What is good enough: a clear demonstration of the product's one interesting property, a hook that names the problem, under fifteen seconds. It does not need to be beautiful. It needs to be legible.
This is exactly where generated video earns its place — not because the output beats a real shoot, but because a real shoot for a product you will drop next Tuesday is not a sensible use of anything.
Scaling — a product that proved out
Now the arithmetic inverts. One product is taking real spend, creative fatigue is the binding constraint, and the quality ceiling matters.
What you need: original footage, real creators, variety that lasts months.
Order the sample. Commission the creators. The cost that was absurd during testing is trivial against the spend the product is now carrying.
The sellers who struggle are usually the ones who never make this transition — still running generated test creative on a product doing $40k a month, wondering why it fatigued.
Making a test ad from supplier stills
A workable sequence when you have three product photos and nothing else:
- Find the one interesting property. Not three features. One thing that makes someone stop. If you cannot name it, the product is probably not a good test candidate.
- Write the hook as a problem, in under ten words, spoken in the first two seconds.
- Generate or assemble three to five frames that tell a tiny story: problem, product, result.
- Add modest motion. Short clips, camera movement rather than object movement — generative models handle a dolly far better than a demonstration.
- Voiceover and captions. Most of the feed is muted; captions are not optional.
- Produce three different openings on the same body. The opening is the variable that matters.
What to avoid asking a generative model for: hands manipulating the product, readable text on packaging, anything over about eight seconds. These are the known weak spots and trying to force them wastes credits.
Reading the test honestly
With fifteen products and small budgets, the temptation is to call winners too early.
- Hook rate first. If under about 20% watch past three seconds, your opening failed and the product was never tested. Fix the opening and rerun before judging the product.
- Enough spend to mean anything. A product judged on $12 of spend has not been judged.
- One variable at a time. New product and new hook style and new landing page tells you nothing.
- Expect most to fail. That is the model working, not a problem to solve.
The margin reality
Dropshipping margins are thin, and video advertising is an expense before it is an asset. Two numbers worth keeping in front of you:
Cost per tested product. Creative plus the test budget. If testing one product costs $80 and one in eight works, each winner costs $640 to find. That number sets how much margin a winner has to carry.
Cost per creative variant on a scaling product. Once something works, you will need a lot of variants. If each costs $200, fatigue becomes expensive fast; if each costs $2, you can refresh weekly.
Generated video mostly changes the second number, and it changes it by an order of magnitude. That is its actual role here.
Where tools fit
For the testing phase, Faysell goes from a supplier product photo or a listing URL to a vertical ad with script, shots and audio. It is built for producing several variants quickly rather than one flagship asset, which is the shape of the testing problem.
Two honest caveats for this use case specifically: supplier sites sometimes block URL imports with anti-bot pages, so uploading the image directly is the reliable path; and the output is generated, so the product is reconstructed rather than copied — for heavily branded packaging, check the label survives before building a campaign on it.
For scaling, order the sample and commission real creators. No tool substitutes for that, and pretending otherwise would cost you more than it saved.
The summary
Do not use supplier video. It is not yours, everyone else is running it, and it is the most common route to a restricted account.
Treat testing and scaling as different creative problems. During testing you need cheap variants fast, and generated video is a good fit. Once a product proves out, order it, film it, and commission creators — the economics flip completely and most people miss the moment they flip.