Skip to main content

What images work best for AI generation

A short guide on how to work with artificial intelligence so it generates the product images you actually want to see.

Vita avatar
Written by Vita
Updated over a month ago

Think of AI like a very talented artist with sharp eyes — but those eyes can only see what you show them. Every image you upload to Assets becomes a brushstroke in your future ads and videos. If the photo is weak, the output will be weak. If it’s strong, AI has everything it needs to surprise you.

Here’s how to give AI the kind of input it loves.


Show the whole product

Half a shoe? AI won’t guess the rest. A cropped bottle? Expect strange results. Always upload the product in full. Clean backgrounds are best, and skip any random text that isn’t part of your product.

👉 Pro tip: When taking a photo, place your product against a neutral background (white, beige, or light gray). It makes the item pop and saves AI from guessing what’s in the shadows.

Quality matters more than filters

If you can’t make out the details like the fabric pattern on a dress, the measurement lines on a bottle, AI won’t either. You can enhance an image in Video Studio, but even AI can’t turn a blurry photo into something worthy of a Vogue cover. Start clear, stay clear.

Look: When a photo is blurry or buried under heavy filters, AI reads it as flawed — and may start “improving” by making things up.

👉 Try this: Zoom in on your image. If it stays clear and the details are easy to spot, you’re likely good to go. If not, a sharper version can help — AI might still work with it, but the result could be less precise. Don’t be afraid to test it, and at the same time, don't get upset if the result is not quite what you expected.

One product, one image

Imagine uploading a photo with five identical caps. Which one should AI pick? Exactly. Keep it simple: one product per frame. That way, the generation knows exactly what to focus on.

Look: When a photo shows two (or more) identical bags that are not the main focus, one is half covered by hands while the other blends into the sofa, AI may start filling in the gaps and imagining details that are not really there.

👉 Shortcut: If you sell a bundle, upload a single photo that shows the items together. Later, when adding the product, you can describe the bundle in text. This way, AI will have one clear image to work with and present it more naturally.

Square, rectangle… doesn’t matter

Good news: AI doesn’t care if your image is square or rectangular. For video, it reshapes everything into 9:16 when you working with Image to video feature. If your asset is horizontal, AI might reposition the product to fit. Don’t have a vertical version? No panic, just give AI more context in the description so it knows what to highlight.

Extra details make results shine

Want realism? Feed AI more than just the picture. Tell it the material such as leather, glass, cotton, and you’ll see the output look closer to the real thing. Think of it as giving AI extra paint colors for its palette.

👉 Try this: Add at least one detail about material or texture (e.g., “glossy glass,” “soft cotton”) to your description for more lifelike images.

Digital products need a twist

Not everything is a bottle or a bag. If you’re promoting digital goods, show them in their natural environment:

  • Templates: Display them together, clearly enough that AI can recognize their structure.

  • Books: Put the text on a phone, an e-reader, or a laptop mockup. Uploading plain text alone may lead the AI to place it on a random background.

The golden rule: input = output

AI doesn’t create miracles; it reflects what you give it. Provide clear images and thoughtful descriptions, and it will produce results that truly represent your product.

✨ That’s it. In the end, these aren’t limitations but shortcuts. Spending a few extra moments to choose the right visuals will save you far more time when editing later.

Did this answer your question?