How to Add Realistic Shadows to Product Photos With AI

A generated background can look convincing until the product appears to float. The usual fix is to ask the AI to regenerate the whole image with better lighting and shadows, but that gives the model another opportunity to change the SKU. For ecommerce, a safer approach is to treat the shadow as a separate part of the composite: keep the photographed product intact, build the shadow around it, then check that the shadow agrees with the scene.

This matters most for products where small changes are commercially important. A bracelet clasp, gemstone setting, logo, bottle label or shoe stitching should not be redrawn just because the image needs more depth.

Start with the product, not the shadow

Before adding any shadow, put the product in its final position and scale. Check perspective, rotation and where it should physically touch the surface. A shadow cannot rescue a product that is already floating two centimetres above the table or sitting at an angle that does not match the background.

If you are working from a cutout, keep the product on its own layer and preserve an untouched copy. That layer is your SKU reference. The shadow should be generated or drawn underneath it, not by asking an image model to reinterpret the whole composition.

Step 1: decide which shadow the image actually needs

Not every product image needs a dramatic cast shadow. For clean catalogue images, a subtle contact shadow is often enough. This is the small darkening directly beneath the places where the object touches a surface. Its job is simply to tell the eye that the product is resting there.

A directional cast shadow does more. It shows where the light is coming from and gives the scene depth. It is useful in lifestyle images, campaign graphics and compositions where the surrounding objects already have visible shadows.

Reflections are different again. A glossy acrylic or polished stone surface may show both a shadow and a reflection. Adding a mirrored copy of the product is not a substitute for a proper shadow, and using a reflection on a matte surface will usually make the scene look less believable.

Step 2: lock the product pixels

The lowest-risk workflow is to isolate the edit. Keep the real product on a protected foreground layer and generate the shadow only in the empty area beneath or around it. If your editor supports masks or selections, constrain generative editing to those background pixels.

Dedicated tools can make this faster. Photoroom's current AI Shadows feature, for example, can generate soft, hard or floating shadows and offers advanced controls such as shadow spread and product orientation on the web app. Photoroom also supports applying AI Shadows to batches of product images and to connected Shopify catalogues. The useful idea is not that one tool is mandatory, but that shadow generation can be separated from full product regeneration.

If an automatic result changes the product edge, creates a dark halo, fills gaps between chain links or paints shadow over the product itself, reject it. The shadow layer is expendable. The approved product layer is not.

Step 3: match the shadow to the light source

Look at the scene before choosing a shadow direction. Where are the brightest highlights? Which side of nearby props is illuminated? If light is clearly coming from the upper left, a strong cast shadow travelling in the same direction will contradict the scene.

The edge quality matters too. A large softbox or overcast window produces softer transitions. A small, hard light source produces a sharper shadow. The shadow should normally be darkest and most defined near a contact point, then become softer as it moves away from the object.

This is where generic prompts such as "add a realistic cinematic shadow" can fail. A shadow is only realistic relative to the lighting that already exists. Describe the actual scene instead: soft window light from the left, product resting flat on a pale stone surface, subtle contact shadow directly underneath, with a soft cast extending slightly to the right.

Step 4: standardise shadows across a catalogue

For a single lifestyle image, small variations may not matter. Across a catalogue, they do. Fifty products photographed against the same background will look inconsistent if some have tight shadows, others have long shadows, and a few appear to float.

Create a simple visual standard for each image type. Keep the same canvas size, product baseline, approximate light direction and shadow softness. Then adjust only what the product's orientation requires. A standing perfume bottle and a bracelet lying flat should not receive identical shadows, but they should still look as though they were photographed in the same studio.

Batch shadow tools can save time here, but review a representative sample before applying the result across hundreds of images. Automation is most useful after the visual rule is proven.

Jewellery needs a stricter check

Jewellery exposes shadow errors quickly because the contact geometry is complicated. A fine chain may touch the surface at many tiny points. A ring standing upright has a very different footprint from a ring lying flat. A raised gemstone setting can create several small areas of occlusion rather than one solid dark blob.

Thin chains are especially easy for automatic tools to misread. A generated shadow can become thicker than the chain itself, bridge open links or make a delicate piece look visually heavier. Transparent or translucent stones should also be checked carefully because a dense opaque shadow may not fit the way light passes through the real material.

Polished metal creates another problem. The product's original reflections already imply a lighting environment. If a silver pendant reflects a bright source from one side while the new background and shadow imply the opposite direction, the composite can feel wrong even when no individual element looks obviously broken.

When AI shadows are the wrong choice

If the original photograph already has a clean, believable real shadow, preserving it may be better than replacing it. For highly reflective, transparent or refractive products, manual compositing or a controlled reshoot can also be more reliable than an automatically generated result.

The goal is not to use AI in every step. It is to remove repetitive work without weakening product accuracy. A plausible shadow that conflicts with the real material, contact points or lighting is still a bad product image.

A fast approval check

Review the finished image twice. First, view it small and ask a simple question: does the product feel grounded in the scene? Then zoom in and compare the product edges with the untouched original. Check fine gaps, transparent areas, labels, clasps, chains and any place where the generated shadow meets the SKU.

For listing images, also check the current rules for the specific marketplace and image slot before publishing. A shadow that works well in a secondary lifestyle image may not be appropriate for every main-image requirement.

If you use Lustra Studio or another AI product-imaging workflow to create new scenes, it helps to separate the approval into two questions: is the product still accurate, and does the environment look physically believable? Shadow quality belongs to the second question and should not require sacrificing the first.

The practical rule

When the product is already correct, do not regenerate it just to fix a shadow. Protect the SKU, generate or adjust the shadow separately, match it to the actual lighting, and standardise the result across the catalogue.

A good shadow should be almost boring. Buyers should notice the product, not the editing. If the shadow quietly makes the object feel grounded and believable while the photographed SKU remains untouched, it has done its job.

Sources: Photoroom Help Center, “How to use AI Shadows”, reviewed 1 September 2026; Photoroom Help Center, “Apply AI Shadows to a batch of images”, reviewed 1 September 2026; Photoroom Help Center, “AI shadows or layer shadows?”, reviewed 1 September 2026.