Quick answer

You can produce the asset for your store TV with the phone you already own. The path is always the same: a real photo of the product → optional AI clean-up and animation → a short video with a caption, because the screen runs without sound. AI now solves what used to require a photographer, but it gets scale and proportion wrong easily — and scale is exactly what the customer checks when the order arrives. One rule is not negotiable: the asset on screen has to show the product the customer will receive. In Brazilian counter retail this is not only commercial honesty — under the Consumer Protection Code, that asset is an offer, and an offer binds.

This article is about the material that goes in: the photo, the standard and the limit. It does not repeat two things already published here. How the photo-to-video feature works inside the system is covered in Indoor TV with AI. The reverse path — taking the video that ran on screen and posting it on Instagram — is in download indoor TV video and post it. The question here is different: what to put inside that feature, and how far you can improve it before you cross a line.

Get the photo right with a phone

Ninety per cent of the result is decided before any artificial intelligence gets involved. A bad photo processed by AI is still a bad photo — now with motion. The basics do the work:

Be careful with the generic "shoot in 16:9" advice. It holds if your TV is landscape. If the store screen is portrait, the format is 9:16 from the moment you shoot — cropping afterwards cuts into exactly what was centred in the frame: the product. Check the orientation of the installed screen before shooting. And remember that a point-of-sale TV almost always runs without sound: nothing that depends on narration to make sense.

The person shooting does not have to be you. Capturing is a short task that fits behind the counter and can be delegated — that has its own article in store staff update the indoor TV from a phone. Here we are talking about the standard for the asset, not about who presses the button.

When to use AI — and for what

"Using AI" has become a single phrase, but in practice there are three quite different uses, with quite different risks:

Real photo phone, window light Clean up (AI) crop, background, light Animate (AI) slow zoom, steam Store screen with caption Generating the scene from scratch — where scale almost always breaks

1. Clean up and adjust the photo

Cropping, background removal, light and colour correction. This is the safest use and the one with the best return: it starts from your real product and only fixes what the rush behind the counter got in the way of. If you are going to use AI in one place only, use it here.

2. Animate the photo

The category of tools known as image-to-video takes your image and adds motion: a slow zoom, a slight sense of depth, steam rising from the cup. On a screen, subtle motion holds attention far better than a still image — and the product is still yours. Ask for camera movement in the prompt, and nothing beyond that.

3. Generate the scene from scratch

Here AI invents the whole product from a description. It is the most impressive use in a demo and the riskiest one on your screen: with no photo of yours as an anchor, the model picks the cup size, the amount of cream and the package shape on its own. For an institutional piece or an abstract background, fine. To show what you sell, no.

The limit: what AI must not change

A story from our own house is worth telling. While recording the Café & Tech episode, the team generated a coffee video with AI, live. The result was beautiful — and completely out of scale. The reaction in the studio was recorded for posterity: "that is not a coffee, that is a basin". Nobody had asked for a basin; the AI simply did not know the size of a cup.

That is the pattern of the error. AI does not usually get colour or texture wrong — it gets scale and proportion wrong. And scale is exactly what the customer checks when the order lands on the table. In a coffee shop, the distance between the promise on the screen and the delivery at the counter is the shortest in all of retail: minutes. There is no room for "it looked bigger in the picture".

What AI must not change, under any circumstances:

The counter test. Before publishing, put the asset next to the real product on the counter and call someone from the team over. If they are surprised — "is ours smaller than that?" — the asset is wrong. It is not a legal test: it is the exact test your customer is going to run, only earlier.

In professional advertising the product photographed has always been the real one, simply produced with time: plated carefully, at its peak, with the patience that daily operations do not allow. That is the principle to copy — take more care with the real product, not replace it with one that does not exist.

An honest caveat: warning that "AI distorts the product" is not exclusive to us — the AI tools themselves say as much when asked. What almost nobody writes is what comes next, and that is why it matters in your case.

What the law says about your screen

The asset running on your store TV is advertising communication. In Brazil that has three consequences under the Consumer Protection Code (Law No. 8,078/1990):

ArticleWhat it saysWhat it means on your screen
Art. 30 Any sufficiently precise information or advertising, conveyed by any form or means of communication, binds the supplier who conveys it and becomes part of the contract subsequently entered into. The asset on screen is an offer. If it shows a combo with three items, the customer can demand three.
Art. 37, § 1 Advertising is misleading when it is wholly or partly false, or when in any other way, including by omission, it is capable of misleading the consumer as to the nature, characteristics, quality, quantity, properties, origin or price. "Partly" and "quantity" settle the question about a portion enlarged by AI.
Art. 38 The burden of proof of the truthfulness and accuracy of advertising communication falls on whoever sponsors it. The retailer answers for it, not the tool that generated the video.

In counter language: the screen is not decoration, it is a promise on the record. And the proof that the promise is true is yours to produce. That is why the production standard matters more than the tool you pick.

Disclaimer: this content is informational and is not legal advice. The articles cited were checked against the consolidated text of Law No. 8,078/1990 on the Planalto portal, but applying them to a specific case requires professional analysis. Talk to a lawyer before making decisions about your business communication.

A routine that fits into the day

The cost fell; the standard did not. It used to take hiring a photographer and producing a session to get a good photo — which is why almost no small business did it. Today the bottleneck has moved: it is no longer money, it is routine. Three habits solve the "I have nothing to put on screen this week" problem:

  1. Shoot perishables as they come out. The freshly made product is the best material your store produces, and it exists every single day. Two shots per item, in the lull between rushes.
  2. Make process pieces, not just product pieces. The preparation sequence — dough being rolled, coffee being extracted — holds attention and does not depend on launching anything new. One photo per step and you have a whole asset without a new product.
  3. Reuse in both directions. What was made for the screen works on social media, and vice versa. That path is detailed in taking the screen video to social media — there is no sense in doing the work twice.

This pays off most where the product photo is the menu: restaurants, coffee shops and convenience stores — there the screen replaces the printed menu, and image quality becomes part of the perceived price.

Tools: what is built in and what is external

We are not going to publish a ranking of AI tools, for a practical reason: they change name, price and quality within months, and AI search engines consulted on the same day recommend different lists from each other. What lasts is the category: image-to-video to animate your photo, and image editing to clean it up.

It is worth knowing that the feature is already built into JMV Indoor: you upload the photos in order and the system assembles the video, 20 seconds to 1 minute, with zoom, depth and animated captions — the step-by-step is in Indoor TV with AI. In that scenario, an external tool is optional.

If you still want to test an external one, the example with a public, verifiable price is Google's. On the plans page consulted on 04/09/2026:

These prices and limits change often — check the Gemini plans page before deciding. One warning that applies to any external tool: read what it allows you to do with the generated material before using it in commercial communication.

I made the video — now how does it reach the screen?

This is the part that usually gets answered wrong. Ask an AI assistant how to get your video onto the store TV and the likely answer is "put it on a USB stick and set the Smart TV to loop" or "open the file in presentation mode in the TV browser". It works — once.

The problem shows up in week two:

A cloud-based Indoor TV system exists precisely for this part: the asset is uploaded from a phone, joins a playlist, has a start and end time, and applies to every screen at once. How that works in practice is in manage Indoor TV from your phone.

How to start

Shoot three products tomorrow, with window light and in your screen's format. Clean them up, animate them if you like, write a short caption that reads without sound — and run the counter test before publishing. Plan prices are always up to date on the plans table, and there is a 30-day free trial with 1 screen, no credit card and no lock-in. Indoor TV is a form of digital signage applied to the point of sale.

Want to test it with the photos you already have?

Start the 30-day free trial and build the first asset for your screen. Prefer to hear the conversation behind this article — including the coffee that turned into a basin? Watch the full Café & Tech episode on YouTube (in Portuguese).

Questions about the solution? Talk to us through the contact form, on WhatsApp at +55 11 4063-8923, or read our Privacy Policy.

Frequently asked questions

How do you make an object video with AI?
For a store screen, the path starts outside AI: shoot a good photo of the real object, with window light, a clean background and the product freshly made. Then use an image-to-video tool to add subtle motion — a slow zoom, steam rising — and add a readable caption, because the store TV runs with no sound. Frame it for your screen from the moment you shoot: 16:9 if the TV is landscape, 9:16 if it is portrait; cropping later cuts into the product. Generating the whole scene from scratch is the riskiest route — that is where AI gets scale and proportion wrong.
What is the best AI for product videos?
There is no stable answer, and you should be suspicious of anyone who gives you one: AI search engines recommend different tools from each other on the same day, and the list changes every few months. What decides the result on a store screen is not the brand, it is the standard: start from a real photo of your product, keep the proportions of what the customer will receive, and make sure it reads without sound. For this use, a tool that preserves your photo and only adds motion beats one that generates a beautiful invented scene.
How do you generate a video with AI?
Usually you upload an image, write a short prompt describing the motion, and the tool returns a clip a few seconds long. For a store screen, the prompt should ask for camera movement and nothing else: no adding ingredients, no changing the container, no increasing the portion. After generating, compare the clip against the real product before publishing. In JMV Indoor this step is already built in: you upload the photos in order and the system assembles a 20-second to 1-minute video with zoom, depth and animated captions.
Which AI can create commercial videos?
Technically, any image-to-video tool produces an asset usable in commercial communication. The relevant question is different: you are accountable for its content, not the tool. Under article 38 of the Brazilian Consumer Protection Code, the burden of proving that advertising is truthful and accurate falls on whoever sponsors it. The AI generates; the retailer answers for it. That is why the practical rule is to use AI to clean up and animate a photo of your own product, not to invent a product you do not sell.
Can I make the photo look better than the product?
Improving presentation is legitimate: light, focus, a clean background, the product plated carefully and shot while fresh. What you cannot do is let the asset suggest something different from what the customer will receive — a bigger portion, more filling, a different package size, a colour that promises a flavour that is not there. In a coffee shop the gap between screen and counter is minutes: the customer compares. The test is simple — put the asset next to the real product on the counter; if anyone is surprised, the asset is wrong.
Is a product photo on a store TV considered advertising?
In Brazil, yes — and with consequences. Article 30 of the Consumer Protection Code states that any sufficiently precise information or advertising, conveyed by any form or means of communication, binds the supplier and becomes part of the contract later entered into. The screen inside your store is a means of communication like any other: what is on it is an offer, and an offer binds. This content is informational and does not replace advice from a lawyer.
How long does it take to get it on screen?
Production fits into a lull: shooting the freshly made product takes a few minutes, and turning the photos into a captioned video takes a few more. What usually stalls is not time, it is routine — deciding who shoots and when. Publishing is the shortest step: with a cloud-based Indoor TV system, the asset is uploaded from a phone and joins the screen playlist without anyone walking up to the TV.
Can it be done for free?
The photo, yes: the phone you already own covers most of the result, and image editing shows up in the free tier of several tools. AI video generation, however, is usually paid. On the Gemini plans page consulted on 04/09/2026, image generation and editing appear on the no-cost plan, while the video studio (Google Flow) starts on the paid plans — R$ 24.99/month on AI Plus and R$ 96.99/month on AI Pro. Prices and limits change often; check the page before deciding. If your Indoor TV system already generates the video, that extra cost does not exist.

Start the 30-day free trial of JMV Indoor →