An ecommerce brand is not short of things to show. It has 40 or 400 products and no plan to turn a catalogue into a posting schedule. The obvious plan, one video per product in catalogue order, fails. Here is what works instead.
What each product needs first#
Product video quality is set by the inputs, not the generator. For each SKU you feature, have four things on file:
- A clean pack shot on white. Straight on, product filling the frame, label readable at full size. This is what lets a presenter hold your actual product instead of a lookalike.
- A shot in use, at human scale. A bottle alone is a shape. A bottle in a hand is a size, and size is what shoppers misjudge most.
- One sentence a competitor cannot say. Not the tagline. The line your best customer uses when they recommend it.
- The one objection. Price, fit, shipping time, “will it work for my skin”, “is it the same as the cheaper one”. This becomes its own video, often the one that sells.
If you cannot write the last two for a product, it is not ready for content. That is a merchandising problem, not a video problem.
Five moments that earn a video#
Plan by moment, not by product. Each SKU passes through some of these, and one script skeleton works across the catalogue with the nouns swapped.
1. It exists#
The launch. One claim, product visible in the first second. A single fade-in suits a hero product where the object is the argument. A product spokesperson suits a claim that needs explaining. For several items at once, a slideshow is faster to make.
2. Somebody doubted it#
The objection video. Answer the sentence that stops purchases. “It’s expensive”: show cost per use. “Will it fit”: measure it against something everyone owns. Few brands make this because it feels defensive. It is not. It answers the reason people leave the cart.
3. It compares#
Side by side with another product, or against the category default. Rankings and tier lists suit this, because disagreement in the comments spreads the video. Do not make comparison claims about a named competitor that you cannot prove.
4. It came back#
Restocks and drops. These have a deadline built in, which most organic content lacks. Timing matters more than polish: a restock video three days late is worse than an average one posted that morning, so schedule it with the restock.
5. The season moved#
The same product reframed for gifting, back to school, summer or your category’s calendar. The script changes and the product shots do not.
Photo rules that decide render quality#
Generated product video comes back to the source image. In order of how much damage they prevent:
- Fill the frame. A product that takes up a small part of the photo gives the model little to work with, and it invents the rest.
- Flat, even light. Hard shadows on the source look wrong once the product is placed in a new scene.
- Label facing the camera. An angled label has to be rebuilt, and rebuilt text is the most common giveaway in AI product footage.
- One product per image. A bundle shot gives you a presenter holding an unclear number of things.
- Full resolution. Not the thumbnail from your storefront.
Test one SKU end to end before you shoot 40. Watch it on a phone and check whether the label survived. If it did not, fix the photo, not the prompt.
Some categories have limits no photo fixes. A generated scene can hold a bottle well but cannot tell a clothes shopper whether something will fit. Show measurements and real fit photos there.
Make the videos#
With Looptake, describe the video in one sentence, add the product photo, or paste the product page URL so the video uses your brand and product details. Looptake writes the hook and script, picks a format (product spokesperson, listicle, ranking, split screen, fade-in, slideshow and others), adds visuals, voiceover, word-timed captions and music, and renders 9:16. Edit any scene before it goes out. For presenter-led videos, see AI UGC; for everything else, the AI video generator.
For a whole catalogue, upload a CSV with one row per video and review the batch.
Plan a quarter, not a catalogue#
An example for 40 SKUs. Five videos for each is 200 near-identical clips, and nobody follows that. Instead:
- 8 hero SKUs get launch, objection and comparison videos: 24 videos.
- The other 32 appear in group content: slideshows, rankings, a range walkthrough: about 12.
- Restocks and seasonal moments: about 8.
That is about 44 pieces over three months, roughly one every other day. A real cadence, not a wish. Mix cheap text formats with presenter videos so the plan fits your volume. Plans differ by volume, so check pricing before you commit.
Why one video per SKU fails#
- The feed becomes a catalogue. Nobody follows a catalogue, so the account has no reason to exist between purchases.
- Weak products get the same effort as strong ones. In most catalogues a few products carry the revenue. Give them the videos.
- You run out. Forty SKUs at one video each is six weeks, then you are back at the start.
Let one SKU appear in a launch, an objection video, a comparison and a seasonal recut across a quarter. Same product, four different reasons to watch.
Never pass off an AI presenter as a customer#
A generated presenter is not a customer and must never be presented as one. A presenter demonstrating a product is fine. A presenter claiming to have used it for 30 days is a fake endorsement, and in health and beauty that is the kind regulators act on. More in AI UGC ads.
Tie videos to orders#
Ecommerce has a transaction, so a video can be linked to revenue, not just views. Give each post its own link to the product page with UTM parameters. Then “which video sold anything” has an answer, and it is often not the one with the most views. Broad, high-view content tends to sell little directly, while a lower-view objection video can drive purchases. Judge by sessions and orders, not views, or you will cut the video that sells.
Set up attribution before the first post. You cannot add it to a quarter of content after the fact. Then let TikTok automation post the approved batch on schedule, and feed what sold into the next batch.