Automatic product videos for new items in your store: template feeds, apps or an AI API

This page is for a store owner or a merchandiser (the person who prepares new products for sale) who adds items on a regular schedule and wants a short video for each one. If you have hundreds of products already in the catalog, start with video for every product when you have hundreds of SKUs (a SKU is one product line you sell; what to film, generate or skip across many products helps you sort them). This page is about the next problem: a new item arrives, and a video should follow with as little hand work as possible.

I will compare four routes, from the cheapest to the most technical. Higgsfield appears in two places, and I will say plainly where it is a poor fit.

This page may contain affiliate links from which Blur Studio earns commissions.

The problem: a video for every new item, every week

Say you add 50 new items a week. Filming each one takes about 20 minutes with setup. That is more than 16 hours a week, so it never happens. The new items go live with photos only, and the catalog slowly gets two kinds of products: ones with video and ones without.

Automation sounds like the answer, but the word hides two different jobs:

• Templated video. A fixed layout takes your product photo, name and price and makes a 5 to 15 second clip. It cannot invent anything, because it only uses what is in your data.
• Generative video. An AI model takes a product photo and makes new motion from it. It is more eye-catching, but the model can change the shape, colour, text or size of the product, and add details that are not real.

For a weekly flow of ordinary catalog items, the first job is usually the right one. Keep the second for your hero products, the few items where a bit of motion is worth a person checking every frame.

Option 0: a spreadsheet and manual uploads

This is the cheapest route and it is enough more often than people expect. Keep one sheet with a row per new item: SKU, product name, photo link, video status, owner and date. Once a week, one person films or produces the videos for the top five items by expected sales, and the rest go live with photos.

The spreadsheet also becomes the backbone of every automatic route below. All of them need clean data: one row per product, one photo link per row, and a status column. If your sheet is messy, no tool will fix that.

Stay here if you add fewer than ten items a week, or if your products are tiny, shiny or see-through (hard cases for any generative tool).

Three ways to automate

Route	What it does	Who sets it up	Main risk
Template feed	Spreadsheet row in, templated clip out	You, in an hour or two	Looks the same for every item
Store app	Makes clips inside your shop and can publish to product pages	You, from the app store	App may break after a theme or platform change
Generative API	Your own code sends a photo to an AI model and gets a clip back	A developer	Product shape changes, running costs, upkeep

Template feeds: how a spreadsheet becomes 50 videos

Creatomate is a service that makes images and videos from templates and a data feed (a table of rows). You design one layout, map the columns (name, price, photo) to it, and it renders one video per row. Its own tutorial on bulk product videos describes this approach and mentions a connector tool called Zapier (a service that links web apps) to make a video when a new product is added. That tutorial is dated 2023 and is written by the vendor, so treat the claim that it scales to thousands as marketing, and check it yourself.

On Creatomate's pricing page on 3 November 2026, the Essential plan showed $45 a month billed yearly ($540), with 2,000 credits and 1,000 feed rows. The page says one image costs 1 credit, and one minute of 720p video costs about 14 credits. It also says rendered files are hosted for up to 30 days, so copy them to your own storage. Prices change, so confirm on the live page.

Here is why this works for new items. The video is made only from your data, so it cannot add a feature that does not exist. The cost is fixed per render. Nothing odd can appear in frame 40. The drawback is that every video looks alike, which is fine for catalog pages and weaker for ads.

Two set-up details decide whether a feed works. First, the sheet: export your products as a CSV file (a plain table), then check that every row has a clean photo link, a short name and a price, because one broken row makes one broken video. Second, the template: decide the shapes you need before you design it, for example square for a feed and tall for Stories or Reels, and keep the logo, font and colours the same on every video so the set looks like one brand. Whatmore, a store video vendor, describes the same idea of saving one creative as a template and applying it across a product sheet, and says it applies to new batches. That is its claim, so test it on three products first.

Store apps: what to check before you install

A store app (a plug-in you install in your shop, for example on Shopify, a popular online store platform) can make clips and publish them to product pages. One example is Provid.AI. Its Shopify listing showed a free plan with 5 videos, Starter at $9.99 a month for 20 videos, Growth at $29.99 for 100 videos with bulk creation and AI voiceover, and Pro at $79.99 for 500 videos. The listing had one review, one star, from January 2026, with a reply from the developer. I am not saying the app is bad. One review tells you nothing either way, which is exactly the point: you have no outside evidence yet.

Before you install any video app, ask:

• How many reviews does it have, and when were the latest ones written?
• Are the videos hosted by the app or stored in your shop? What happens to the videos on your product pages if you uninstall the app or stop paying?
• Can you set the review step to Draft, so nothing publishes without a person approving it?
• Can you test on three products on a free plan first?

Where Higgsfield fits: the API is a developer product

Higgsfield, the platform for AI images and video, sells its models in two separate ways. The first is the web app, paid for with a monthly plan and credits (Higgsfield's unit of spend). The second is the Higgsfield API. An API (application programming interface) is a way for one program to talk to another. Higgsfield's help center article, dated 16 September 2026, says the API is a separate product from a plan. Here is what that article says:

• It runs on a prepaid US dollar balance. The minimum top-up is $5 and funds expire one year after they are added.
• Your plan, its credits and its Unlimited modes do not apply to the API. They exist only on the website.
• Creating a key gives you 20 concurrent requests (20 jobs running at once).
• Video is priced per second and images per image. Failed generations are not billed. An estimate call returns the cost before you run a job.
• Jobs run in the background; you check status by polling (asking again every few seconds) or by a webhook (a message the service sends to your system when the job finishes).
• Outputs are kept for at least 7 days and may then be removed. Section 4.4 of the Terms says Higgsfield does not restrict commercial use of Outputs.
• The same section lets Higgsfield use inputs and outputs to improve its models, except under Enterprise agreements. Unreleased product photos are inputs, so decide whether that is acceptable. Section 11.7 also says you must not present outputs as human-made and you are responsible for any AI disclosures the law requires.

What this means in practice: a store owner who wants automatic videos needs someone to write the glue. A store event fires when a new product is added, your code calls the API with the photo, waits, downloads the result within 7 days, and uploads it to your shop. Nothing on Higgsfield's API pages describes a no-code store connection. I could not read per-model prices as text on Higgsfield's model catalog at open.higgsfield.ai/explore, so I am not quoting any. Read them there and use the estimate call.

My honest advice: for most small stores, the template feed is the less risky first step. The API makes sense if you already have a developer, your hero products justify motion, and you will build the approval step described below.

Use the Higgsfield web app for the pilot, before any automation

The web app is a better place to learn what generative video does to your products than an API project is. In the web app you can upload a product photo and make a short clip, or use its ad tools that start from a product page link (see Higgsfield Marketing Studio and turning product photos into short videos). Higgsfield has not published a credit price for those tools; the price shows on the Generate button, so measure what an approved clip costs instead of guessing. Plan prices are on the pricing explainer, and you should check the live pricing page.

The aim of this pilot is not output. It is to learn, on ten of your own products, how often the first try is good enough to publish. That number decides whether automation is worth building at all.

Cost per video: an illustrative calculation you can redo

These numbers are made up to show the method. They use another company's public prices as a yardstick, not Higgsfield's. On 3 November 2026, the pricing page of fal (a marketplace that sells AI models by usage) listed Kling Video v3 Pro at $0.14 per second and MiniMax H3 Max Turbo at $0.015 per second (Kling and MiniMax are two makers of AI video models). Prices change, so re-check.

Step	Higher-priced model	Lower-priced model
Price per second	$0.14	$0.015
One 5 second clip, first try	$0.70	$0.075
50 new items, first tries	$35.00	$3.75
Assume 3 tries per approved clip	$105.00 per week	$11.25 per week
Cost per approved clip	$2.10	$0.225

Two lessons. First, the keeper rate matters more than the price per second: if only 1 clip in 5 is good, your real cost is five times the price you saw. Second, the cheap model may not look good enough for your products, so the price alone settles nothing. The same cost-per-approved-result method works for the web app; see how to spend Higgsfield credits wisely.

The approval gate: never publish straight from the generator

This is the part most guides skip. Build the check into the flow, not after it.

1. Every generated video lands in a Pending folder or a sheet column with status Draft.
2. One named person changes it to Approved or Rejected. Rejected items go back for another try or fall back to photos.
3. Only Approved videos publish to the product page. The generator never has write access to the live shop.
4. Keep the source photo link, the model name and the prompt in the sheet next to the SKU, so a bad video can be redone.
5. Keep a rollback: you can delete the video and keep the photo for each product.

Name files by SKU. It sounds dull and saves hours the first time something needs fixing.

Product video review checklist

• Shape, colour and proportions match the real product photo.
• No extra logos, no garbled text, no invented labels.
• No feature the product does not have. Higgsfield's own published build log of an AI-built website says the agent invented plausible product specs, so this failure is real, not theoretical.
• Hands, fingers and moving parts look natural.
• Motion does not suggest a wrong size or function (a pocket gadget that looks table-sized, for example).
• No claim is added that you have not approved.
• The shape fits where it will run. A product page, a Reel and a Short want different frame shapes (wide, square or tall), so make one template per placement and check the crop.
• Where the video shows a person or voice, you have the rights and consent for it. Check local law and the platform's current rules.

A 10 item pilot with pass or fail rules

This is a plan you can run, with thresholds I suggest. Adjust them to your margins.

1. Pick 10 products across your 3 hardest categories: reflective, very small, and see-through.
2. Make 3 versions of each, using the same settings.
3. Score each version pass or fail on the checklist above.
4. For each category, count how many products got at least one pass on the first try.
5. Automate only categories where at least 7 of 10 pass first time. Keep the others on templates or real footage.

Record the cost of the whole pilot and divide by the approved clips. That is your cost per approved video, and it is the number to compare with the template feed.

Who should skip the generative route

• You add fewer than ten items a week. A person with a phone is cheaper and safer.
• Your product must be shown accurately: size, fit, texture, working parts. Real footage wins. If a claim has to be real, show the real product.
• You have no one who can approve each video. Without the gate, I would not automate at all.
• You want a no-code connection to your shop. Higgsfield's API pages describe no such connection; a template feed or a store app is closer.
• You need a UGC (user generated content) style ad, one that looks like a customer filmed it, with an avatar made from a product link, with a ready interface and no developer. Creatify (a service that makes avatar-style ads from a product link) is built for that. Its pricing page on 3 November 2026 showed Standard at $10 a month and Pro at $38 a month on annual billing (struck-through list prices of $15 and $55), so monthly billing costs more. Standard had 55 credits and one generation at a time; Pro had 280 credits and up to 5 in parallel. Confirm on the live page.

Also remember outputs from any AI tool may not be unique, and no tool removes your duty to follow platform rules on labelling AI content.

Choosing: a decision table

Your situation	Start with
Under 10 new items a week	Spreadsheet and phone footage for top sellers
Same layout for every item is fine	Template feed (Creatomate or similar)
You want it inside your shop, no developer	A store app, after the checks above
UGC style ads from a product link	Creatify or Higgsfield web tools, with a pilot
Hero products with motion, developer on hand	Generative API behind an approval gate
Accuracy matters most	Real footage

For the wider comparison of tools, see best AI ad generators for ecommerce, and for fixed product lists see a weekly ad refresh. For a build-it-yourself view of the API, see the video generation API page.

How I'd approach this

I would split the flow in two before touching any tool. New items get a templated clip from the spreadsheet, so every item has something. The top five items of the week get real or carefully checked generative video. Then I would run the 10 item pilot on the Higgsfield web app or another generator, and only talk about an API once I knew my approval rate.

If you want this to run on its own, Blur Studio builds the repeatable workflow around these tools: the sheet, the approval step and a shared asset library, so nothing publishes without a check. See how I work with business owners.