ESStart a project

AI product photos and films for brands: how it works and the rules

Old rule: You need a photoshoot for every launch.You need one sharp product photo. AI does the scenes.

How brands turn one product photo into campaign images and films with AI, what AI still gets wrong, and the labelling, copyright and advertising rules to follow.

Facts checked: 4 min read

A brand used to need a shoot for every launch, season and channel: a location, a crew, props, a week of edits, and a reshoot when the brief changed. Now one sharp product photo can become a whole campaign.

We’ve made AI campaigns for Posh Gin, MAGA Beer, Bonus Health, Smash Tennis and WSL. This is how the process works, where AI still falls short, and the rules you need to follow.

How it works

  1. The product photo. One or a few sharp, evenly lit shots of the real product, front and three-quarter view. This is the one thing AI must not invent.
  2. The brief. Where the product lives, the light, the mood, the season, the formats (feed, story, banner, print). Written, with references.
  3. Still images. An image model places the product in each scene. We generate many options and pick a few. Most of the craft is in the picking.
  4. Fixing the product. The label, the logo and the bottle shape are checked against the real photo and corrected by hand where the model drifted.
  5. Film. A video model animates the best stills into short shots: a pour, a pan, a breeze. Then a normal edit: cuts, sound, colour, titles.
  6. Every format. The same scene reframed for each channel, instead of one crop that fits nothing.

The first images usually arrive within 48 hours of the brief. A short film takes days, not weeks.

What AI is good at, and what it isn’t

Good at Still weak at
Places you can’t afford to shoot: a desert at dawn, a yacht deck, a bar in 1960s Havana Small text on packaging: labels drift and need fixing by hand
Seasons on demand: summer in November, snow in July Faces: lookalikes, consent and the uncanny look
Many variations of one idea for testing ads Liquids and physics: pours, splashes and ice can look wrong
Every format from one scene Exact fabrics and textures in close-up
Changing the brief on Tuesday without a reshoot Keeping one character identical across many shots

The fix for most weaknesses is the same: a person checks every frame against the real product and throws out anything that isn’t right.

The rules to follow

Don’t misrepresent the product

AI can make a product look like anything. An ad can’t. If the image shows a different colour, size, finish or result than what the customer gets, it misleads them. In the US the FTC has said plainly that using AI tools to trick or mislead people is illegal. In Operation AI Comply it went after businesses that did. Its rule on fake reviews also bans AI-generated testimonials from people who don’t exist.

Simple test: would a customer who received the product feel the image lied? If yes, don’t use it.

Label it where the platform asks

  • Meta (Instagram, Facebook, Threads): asks people to use its disclosure tool when they post photorealistic video or realistic audio that was digitally created or altered, and shows an AI info label on content it detects as AI-made.
  • TikTok: requires creators to label AI-generated or significantly edited content that shows realistic scenes or people.
  • YouTube: requires disclosure when content shows a realistic scene that didn’t happen, alters real footage, or makes a real person seem to say or do something they didn’t. Clearly unrealistic content and minor edits don’t need it.

We label AI content when we post it, even when a rule might not strictly require it. Audiences don’t punish brands for AI; they punish brands for hiding it.

The EU AI Act

Since 2 August 2026, Article 50 of the EU AI Act requires anyone who uses AI to make realistic generated or manipulated images, audio or video (deepfakes) to disclose it. For clearly artistic, creative or fictional work, a light mention is enough, as long as it doesn’t spoil the work. The 2026 AI omnibus delayed other parts of the Act, not this duty. The only Article 50 relief is a short transition, to 2 December 2026, for the machine-readable marking that AI tools themselves must add.

If you advertise in Europe, treat disclosure as required for realistic AI ads.

  • Your contract decides who may use the work. Ours gives the client full commercial use of everything we make.
  • Copyright protection is narrower. The US Copyright Office’s January 2025 report says that, with today’s tools, prompts alone don’t make someone the author of the output. Human work around it, like selection, arrangement and editing, can be protected case by case.
  • Use tools whose terms allow commercial use, and keep a record of which tool made what.
  • Never imitate a real person, a celebrity or another brand’s look. That’s a legal problem and a trust problem.

Where to start

Pick one product and one channel. Send a sharp photo of the product, a short brief and three images you like. You’ll see within days whether AI content works for your brand, without paying for a full campaign first.

Questions

Can AI photograph my actual product?

It works from real photos of it. You send one or a few sharp, well-lit shots of the product, and the AI builds new scenes around it. The product itself must stay accurate: same shape, colour, label and size. Anything that changes how the product really looks belongs in the bin, not in an ad.

Do I have to label AI images and videos?

Often, yes. Meta, TikTok and YouTube ask creators to disclose realistic AI-generated or altered content, and Meta adds an AI info label when it detects it. In the EU, the AI Act's transparency rules have applied since 2 August 2026: realistic generated or manipulated images, audio and video (deepfakes) must be disclosed. Labelling is cheap; getting caught out is not.

Who owns AI-generated images?

Between you and us, you do: our contracts give the client the right to use everything commercially. Copyright is a separate question. The US Copyright Office says prompts alone don't make you the author, while human selection, arrangement and editing can be protected. Keep your real product photos, logo and packaging: those stay fully yours.

Is AI content cheaper than a photoshoot?

Usually, for volume and variety: dozens of scenes, seasons and formats from one product photo, with no travel, sets or reshoots. A shoot still wins for real people, real places and anything where exact detail matters, like a close-up of a textile. Many brands now do one small shoot for hero shots and AI for everything around it.

Do you generate people?

We avoid faces. Hands holding a product work well; faces bring lookalike risks, consent questions and the uncanny look that makes viewers distrust the whole image. When a campaign needs people, we use real ones.

Sources

  1. EU AI Act Service Desk: Article 50, transparency obligations
  2. European Commission: AI Act, regulatory framework and timeline
  3. Meta: Labeling AI-generated images on Facebook, Instagram and Threads
  4. Meta: Our approach to labeling AI-generated content and manipulated media
  5. TikTok Community Guidelines: Integrity and authenticity
  6. YouTube Help: Disclosing altered or synthetic content
  7. US Copyright Office: Copyright and Artificial Intelligence, Part 2: Copyrightability (January 2025)
  8. FTC: Crackdown on deceptive AI claims and schemes (Operation AI Comply)
  9. FTC: Final rule banning fake reviews and testimonials