Inserting Products Into Existing Photos With AI: What Actually Works

Fstoppers Original

One of the more practical uses I've found for AI in photography is product placement — specifically, dropping a product into a photo you've already taken. Not generating a scene from scratch, but salvaging or extending a shoot you already have.

Here's the scenario: you have a lifestyle shot with a model that would be perfect for a campaign, except the product isn't in it. Traditionally, you'd reshoot the product to match the angle and lighting, then composite it in Photoshop. With a tool like Nano Banana Pro, you can skip most of that and have the AI handle placement, scale, and lighting in one pass.

You don't even need a reference image of the product. You can name it — an old fashioned, an Aperol Spritz, a generic amber bottle — and the AI will build one into the scene.

Woman in white shirt and straw hat sitting at pool edge overlooking coastal cliffs and sea.

How the Tools Compare

I tested this with Nano Banana Pro and ChatGPT 5 Image Generation Pro, asking both to place a glass of Aperol Spritz on the ledge of a pool in an existing shot. Both are capable, but Nano Banana Pro consistently produced a more natural result: better contact shadows, more believable reflections on the glass, and a cleaner integration when I asked it to add a bottle alongside the drink. ChatGPT's version was usable, but the glass read more as pasted-in than belonging to the scene.

Woman in white dress and straw hat sitting on edge of hot spring pool with bottle beside her.

Two Tips That Actually Help

Describe size changes as percentages. If the glass feels too big or the bottle too small, don't say "a little smaller." Say "reduce by 15%." This is the single most reliable way I've found to get predictable scale adjustments.

Treat the label as a separate job. Text on product labels almost always warps. It will usually say the right words, but the kerning, curvature, or clarity will be slightly off — enough that a viewer will sense something is wrong without knowing why. Either composite the real label in afterward in Photoshop, or accept that each re-prompt to fix the label tends to break something else in the frame. In my testing, by the time the label was correct, a shadow or a reflection had gone sideways.

Side-by-side comparison of AI-enhanced and original beach scene with bottles and hands on sand.

Campaign Consistency Is Still Hard

Where things got genuinely difficult was asking the AI to carry the product across multiple shots, the way you would for an ad campaign with varied poses and angles. Consistency drifted from frame to frame. In one attempt, I switched from the cocktail to a pre-mixed bottle, and the AI returned a completely different composition. In another, it silently changed the crop and produced something unintentionally funny — I'll admit I laughed out loud at one result.

Re-cropping to match the original reference hides most of the damage, but it's a reminder that you're directing a collaborator that doesn't quite understand continuity the way a photographer does.

Orange beverage bottles arranged on sand near a person's bare feet at the beach.

The Verdict

You can get a usable lifestyle product shot this way, provided you have patience and a few rounds of prompting in you. Generating a scene from scratch with the product baked in tends to be smoother than inserting a product into an existing photo, but both are viable.

The real value isn't replacing product photography — it's insurance. If you wrap a campaign and realize the product wasn't in a key frame, that's no longer a reshoot. It's a prompt. And to the untrained eye, the difference is getting harder to spot.

Related Articles

No comments yet