You no longer need a product sample, a studio, or a photographer to produce a photorealistic product visual. Here’s an AI product mockup workflow that actually holds up in practice — not just in a demo reel.
01. The Two-Step Approach That Works
The mistake most designers make when they first try this is asking AI to generate the background and the product in a single image. The result almost never matches your actual design — proportions drift, materials look approximate, and the product reads as generic rather than yours. The fix is to split the job in two: generate the environment with AI, then composite your real design into it.
Generate the environment — focus entirely on setting, lighting, and mood in Midjourney, with no product in frame. A working prompt: product photography studio, white marble surface, single window light from left, warm ambient, Canon 5D shallow depth --ar 3:2 --style raw. Generate four to eight variations and pick the strongest lighting setup, not the prettiest single frame.
Composite with Generative Fill — import the chosen environment into Photoshop. Select where the product should sit, clean the selection with Generative Fill, then paste your actual product render or vector on its own layer. Use Camera Raw to match lighting between layers, and add a shadow layer at 15–25% opacity in multiply blend mode.
The underlying logic is worth internalizing rather than just following as a recipe: AI is excellent at generating plausible environments because it has seen millions of them, but it has never seen your specific, unreleased product, so asking it to render that product accurately is asking it to guess. Compositing removes the guess entirely — your product stays exactly as designed, and only the world around it is generated.
This also solves a practical problem that pure AI-generated product shots create constantly: revision requests. If a client asks to see the same product in a different color or with a different logo placement, an AI-generated composite image gives you no separate layer to adjust — you’d have to regenerate from scratch and hope for consistency. A composited file keeps your product as an editable layer, so a color change is a five-minute edit instead of a re-roll of the dice.
02. Getting the Lighting to Match
The single most important technical step in any AI product mockup is light direction matching. Every light source in your generated environment casts shadows in a specific direction, at a specific softness. Your composited product needs shadows facing that same direction and matching that same softness, or the image reads as obviously fake within about half a second of looking at it — human perception is unreasonably good at spotting mismatched shadow logic.
Reads as real
Shadow direction, softness, and color temperature on the product match the environment’s light source exactly.
Reads as fake
Product lit flat or from a different angle than the background; shadow missing, too sharp, or facing the wrong way.
Photoshop’s Neural Filters lighting effects are a reasonable starting point for this, but they rarely nail it on the first pass — plan on a manual refinement step where you nudge shadow angle and softness by eye until it stops looking composited and starts looking photographed.
💡 Pro tip — Before compositing, place a small gray sphere or cube reference into your generated environment image (even just mentally) to read where the light is actually coming from. It’s much easier to match shadow direction against a simple reference shape than to eyeball it directly off a complex background.
03. Where Designers Skip a Step
The most common shortcut, and the one that undermines an otherwise solid workflow, is skipping the shadow layer entirely because the composite already “looks fine” without it. A product floating with no contact shadow reads as fake even when every other element is well matched — the eye expects an object resting on a surface to touch that surface visually, and a soft shadow is what sells that contact more than any other single detail.
The second common shortcut is generating only one or two environment variations instead of the recommended four to eight. More variations cost a few extra minutes and consistently surface a better lighting setup than the first attempt — treat the generation step as cheap and the selection step as the one worth spending real attention on.
Closing thoughts
An AI product mockup workflow that works isn’t about better prompts alone — it’s about respecting the boundary between what AI generates well (plausible environments) and what needs to stay real (your actual product). Split the job at that boundary, match the light carefully, and the result holds up next to an actual studio shoot.
Design Daily Life · Notes on design, daily