For consistent product photos with clean backgrounds and natural shadows, use a background-removal and shadow tool with batch processing, such as Photoroom, Adobe Firefly or Canva, and drive it from a written photo spec sheet: fixed canvas, margins, light direction and background value. The spec sheet is what keeps a catalogue consistent.
Most tool roundups rank tools on one image at a time, and almost any of them can make one picture look good. The harder problem for an online shop is picture 214 matching picture 3, three months later, when the product is a different shape and you are in a hurry. So this tutorial treats consistency as the thing you design first and the tool as the thing that carries it out. It covers the photo spec sheet, shadows that look natural, batch settings, marketplace rules, and a pass-or-fail check before upload. It does not repeat the cost argument against hiring a photographer; that is in AI product photos versus a photographer.
Write the spec sheet before you open any tool
A spec sheet is half a page that says what every image in the catalogue must look like. Without one, each new product gets edited by feel, and slight differences in margin, shadow and white add up until the shop page looks like a car boot sale. Here is a filled-in example for a shop that sells ceramic homeware, which you can copy and adjust.
| Item | Setting | Why |
|---|---|---|
| Canvas | 2,000 by 2,000 px, square | Fits grids and meets marketplace zoom minimums |
| Product size | Longest edge fills 80 per cent of the canvas | Keeps items looking the same size in a grid |
| Margins | At least 8 per cent clear on every side, shadow included | Stops crops from clipping the shadow |
| Background | Pure white (255, 255, 255) for main images; warm grey (#EDEAE4) for lifestyle set | One value per image type |
| Light direction | From upper left, always | Shadows fall to lower right on every image |
| Shadow | Soft contact shadow plus faint ground shadow | Grounds the product without a heavy drop shadow |
| Angles per product | Front, three-quarter, detail, scale shot | Same shot list for every product |
| File | JPEG, sRGB, named brand-sku-angle | Searchable and uploadable in bulk |
Take one hero product, finish it to this sheet by hand, and keep the result as the reference image. Every later image is judged against it, side by side, on the same screen.
Shoot so that the AI has less to fix
An AI background tool starts with a cut-out. The cleaner the edge, the better everything after it. A few habits do more than any setting.
- Use the same light for every product. A window with a white card on the opposite side gives soft, even light that is the same on day one and day ninety. Mixed lamps give colour casts that the tool will happily keep.
- Contrast with the background. Shoot a white mug on a mid-grey sheet and a black item on a pale one. Cut-outs fail where product and backdrop share a tone.
- Keep the camera at the same height and distance for each angle in the shot list, so products share a perspective.
- Include a grey card or colour chart in one shot per session so colour can be matched later.
- Shoot a little wider than the final crop. The tool needs room to build a shadow, and cropping tight loses it.
Backgrounds: white, scene or both
Plan two image types. A white or neutral background for the main image, and a scene or lifestyle image for the second, third and later positions. They have different rules and different risks.
Clean white backgrounds
The task is simple and the tools are good at it. The risk is that "white" varies. Check the background value is actually 255, 255, 255 and not 252 or a faint grey caused by a gradient. Marketplaces that insist on pure white often reject anything else automatically. Remove backgrounds, then add a flat white layer, rather than lightening the original photo, which can clip your product's highlights.
AI-generated scenes
Scene generation places the product on a marble counter, a wooden table or a bathroom shelf, built from a short prompt. It is where tools differ most and where most consistency problems arise, because each generation is a new scene. Three ways to keep a set coherent:
- Fix the scene prompt. Write one sentence and reuse it word for word: "warm oak tabletop, soft morning light from the upper left, blurred cream wall behind, shallow depth of field". Change only the product.
- Use a template or saved style. Several tools let you save a scene or style and apply it across images. Where one is offered, use it instead of retyping.
- Generate the scene once, place products in it. Create one backdrop you like, then composite each cut-out onto it with a matching shadow. Seen side by side, products share an identical background, which is the strongest consistency you can get.
An illustrative sample prompt and what to fix in the result:
Place this ceramic jug on a warm oak tabletop. Soft window light from the upper left, a blurred cream wall behind, shallow depth of field. Keep the jug exactly as photographed: same shape, colour, glaze and logo. Do not add other objects.A typical output has a pleasing scene and two problems: the jug's glaze looks a shade glossier than the original, and the handle's shadow falls to the left, contradicting the upper-left light. Fix the first by blending the original product pixels back over the generated ones, and the second by regenerating or adding the shadow yourself. Always zoom to 100 per cent and compare the product against the source photo, because generation can alter text, labels, stitching and fine patterns without any warning.
Shadows that look natural
A missing or wrong shadow is the most common reason AI product images look pasted on. Three kinds of shadow cover almost every product.
- Contact shadow: a dark, tight line where the product touches the surface. It tells the eye the object has weight. Without it the product floats.
- Cast shadow: a softer, longer shadow running away from the light. Its direction must match the light in the scene and in every other image in the set.
- Reflection: a faint mirrored image on a glossy surface. Use it for glass, polished metal or acrylic displays, and only if you do it for the whole set.
Photoroom, for example, lists AI Shadows on its paid plans, with presets on all of them and adjustable shadows counted separately against plan limits. Look for the same in other tools: a preset for consistency and a manual control for awkward products. Settings to compare are softness, distance, opacity and direction. Choose values once, write them on the spec sheet, and treat them like paint colours: same code every time.
Checks for a believable shadow
- Does the shadow start where the product touches the ground, with no gap?
- Does it fall the same way as in the reference image?
- Is it softer at the far end than at the contact point?
- Is it faint enough to disappear on a white main image but still visible?
- For tall or oddly shaped products, does the shape of the shadow match the product's silhouette, such as a handle or a spout?
Batch processing and what it costs
Doing 300 images one at a time is how a Saturday disappears. Most tools in this space offer some form of batch: upload a folder, apply the same background, size and shadow, download the results. Check the batch rules in the vendor's own terms, because limits on number of images, resolution and number of exports vary by plan.
As a data point for budgeting, Photoroom's pricing page as checked in October 2026 lists Pro at $12.99 a month ($7.50 a month billed yearly), Max at $34.99 ($20.99 yearly) and Ultra at $99 ($82.50 yearly). The plans differ in AI credits and exports: Pro includes 8,000 AI credits and 1,000 exports a month, Max 25,000 credits and 3,000 exports, and Ultra 75,000 credits and 10,000 exports, with 4K images only on Ultra. Batch processing is listed as a separate product, so confirm it is covered before you buy. Prices and limits change, so treat these as what to compare against, not a quote.
Here is a worked sum, with figures marked as illustrative. An online shop has 120 products, four images each, so 480 images, plus 40 new products a month after launch.
- Launch month: 480 images through the pipeline. On a plan with 1,000 exports a month, that is half the allowance, but check how many AI credits one scene generation uses, because scene images use more than plain cut-outs.
- Following months: 40 products at four images is 160 images, comfortably inside the same plan.
- Cost at the lowest paid tier on monthly billing is $12.99, so launch month works out at about 3 cents an image, ignoring your time.
- Time: 5 minutes of review per product for four images is about 3.3 hours at launch, 40 minutes a month after.
The review time is the real cost, and it is the part tools do not remove. For the pros and cons of replacing a photographer entirely, see the comparison linked earlier.
Marketplace and platform image rules
Every platform has an image policy and each differs. Before you replace images across a catalogue, read the current rules for each place you sell. As a typical pattern to check for, marketplace main images commonly require:
- A plain white background, often specified as pure white (255, 255, 255).
- The product filling most of the frame, commonly 85 per cent or more.
- A minimum size, often 1,000 pixels on the longest side to enable zoom, with larger recommended.
- No added text, logos, borders or watermarks on the main image.
- An accurate representation of the actual product.
These are common requirements drawn from seller guidance, and Amazon's own help page is the authority for Amazon, so check it for the figures that apply to your category. Treat any generated scene as suitable for additional image slots, and check whether the platform requires you to disclose edited or generated imagery. Accuracy matters most of all: a customer who receives a product that looks different from the AI-generated picture returns it, and that costs more than any photographer. Online shop AI mistakes that hurt trust and conversions lists the ones that cost sales.
Products that break the tools
Every catalogue has a few items that a cut-out tool handles badly. Spot them early and plan a different route, instead of discovering them in the grid check.
- Glass and clear acrylic. The tool cannot tell glass from background, so it either keeps the old background inside the glass or cuts a hole. Shoot these against a dark backdrop, then composite, or shoot a clean white version by hand.
- Fine edges. Fur, fringe, lace, chains and wire baskets lose detail at the edge. Zoom in, and keep a hand-masked version for the hero image.
- Reflective metal. It mirrors the room, so a tool may treat the reflection as background. Use a light tent and keep the reflections you want.
- White on white. A white product on a white sheet has no edge to find. Use a grey sheet for shooting and add the white background afterwards.
- Products shown with scale. A hand or a coin in the shot is part of the information. Keep an unedited scale shot in every set.
Keep a short list of these products on the spec sheet with the route you chose for each. When the next glass vase arrives, the decision is already made.
A pass-or-fail check before upload
Put this on one page and run it on each batch before upload. Pull 10 per cent of the images at random, plus every image of any new product type.
- Edge. Zoom to 100 per cent along the whole outline. No fringe, no missing handle, no lost chain or strap.
- Detail. Labels, stitching, text and logos on the product match the original photo.
- Colour. Compare against the physical product in daylight. The shade should match what the customer will receive.
- Shadow. Direction, softness and contact match the reference.
- Size and margin. Canvas, fill percentage and margins match the spec sheet.
- Background value. Eyedrop the corner on a main image: 255, 255, 255.
- Set view. Place six images from the batch in a grid. Nothing should stand out.
- File. Name, format and size are correct.
A realistic failure, illustrated. Imagine a shop that ran a candle range through a batch tool with an auto-shadow setting. Every image passed a quick glance. In the grid view two weeks later, the tall pillar candles had shadows twice as long as the jar candles because the tool scaled shadow to product height. The fix was one fixed shadow preset in the spec sheet and re-running 36 images. A grid check at the end of the first batch would have caught it on day one.
Where each type of tool fits
| Tool type | Strength | Watch for |
|---|---|---|
| Dedicated product-photo apps (such as Photoroom) | Fast cut-outs, shadow presets, batch and templates built for shop images | Plan limits on credits and exports; batch may be separate |
| Design suites (such as Canva) | Background remover plus layout, brand kit and resize in one place; AI features draw on a monthly allowance | Allowances per plan; shadow control is basic |
| Adobe Firefly and Photoshop | Precise editing and fine shadow control | Slower per image; partner models in Firefly differ in how they are licensed |
| General chat assistants with image editing | Flexible prompts for scene ideas; good for one-off hero images | Inconsistent between runs; may change product details; weak for batch |
| Your platform's own tools | Already connected to your listings | Usually fewer controls over shadow and style |
For most small shops the answer is one dedicated tool for the batch work and a general assistant for occasional scene ideas, not a stack of five subscriptions. Run the same 10 products through two tools, judge them against your spec sheet, and keep the one whose edges and shadows need the fewest fixes.
Three small shops, three set-ups
A signage company's online shop. Imagine a signage company selling stock signs and brackets online. Products are flat, rectangular and reflective. A white main image with a fixed soft shadow suffices; the risk is glare on laminated signs, which an AI cut-out will not fix. Solve that with the lighting, not the tool.
An interior designer's online homeware range. Imagine a small range of cushions, vases and throws. Scenes matter more than white cut-outs because customers want to see the product in a room. One generated backdrop with consistent light, plus true-to-colour cut-outs, carries the whole set. Check fabric colour obsessively, because textiles are the category where colour returns concentrate.
A solar installer's accessories page. Picture a small online shop selling mounting kits and cable accessories. These are technical items where accuracy beats atmosphere: white backgrounds, a scale shot and a detail shot each. AI scenes would add nothing but risk.
When the stills are done, short video is the natural next step; which AI video tools are worth paying for covers that, and writing the product description with AI covers the words beside the pictures. If you want help setting up the whole workflow around your own catalogue, that is what my AI implementation consultation is for, but the spec sheet above is the part that does the work.
Questions online shop owners ask about AI product photos
Can I use AI backgrounds on marketplace main images?
Usually the main image has strict rules, commonly a plain white background and an accurate view of the real product, while lifestyle scenes are allowed in the additional images. Rules differ by marketplace and category and they change, so read the current image guide before you replace anything.
Why do my AI shadows look fake?
Usually the shadow's direction, softness or contact point does not match the product. A product sitting on a surface needs a dark, tight contact shadow under it and a softer, longer one beyond. Pick one light direction in your spec sheet and use it for every image.
Do I still need to photograph the product well?
Yes. Sharp focus, accurate colour and even light in the original decide the result. AI can replace a background, but it cannot rescue blur, a colour cast or a clipped highlight, and it may alter fine detail such as stitching, labels or text on the product.
How do I keep a new product matching the old ones?
Save the settings and a reference image for the set: canvas size, margins, background value, shadow preset and crop ratio. Re-run those for every new item and compare it beside two existing images before upload.
Further reads
- How to Create Product Photos With AI on a Small Budget — Start with a small budget and a phone camera.
- AI Model Shots for Boutiques: Try-On Images Without a Photoshoot — Try-on images for clothing without a shoot.
- Can You Legally Use AI-Generated Images in Your Marketing? — What you can legally do with generated images in marketing.
- Is Midjourney Worth It for Small Business Marketing Images? — Whether a scene generator earns its fee.
- Taking a Local Shop Online With AI: Product Listings in a Weekend — Get a whole shop online with listings in a weekend.
- AI Jewellery Model Generator: Keep Rings and Chains True to Scale — A step-by-step scale test for AI jewellery models, using a known ring size and chain length so you catch oversized rings and floating chains.
- AI Mirror for a Clothing Store: Is a Smart Fitting Room Worth It? — A small-store payback sum for an AI smart mirror, the vendor claims to test, camera privacy questions and cheaper ways to get the benefit.
- ChatGPT Prompts for Jewellery Shops: 8 for Listings and Clients — Eight prompts for listings, repairs and client messages, each fed from a spec block so every weight, grade and origin traces to your paperwork.
- How to Make AI Ads for a Local Shop: One Ad Set From Start to Finish — One image ad and one short video ad for a local shop, built with AI from a real photo, checked against platform rules and tested on $84.
- How to Use AI to Stage Furniture for Product Page Room Sets — Build room sets around your real product photos, check scale, colour and construction against the piece, and label the result honestly.
- Photography Copyright and AI: Ownership and the Contract Clause to Use — Where AI editing ends and generating begins for copyright, what tool terms say, and contract clauses a photographer can adapt and have a solicitor check.
- Can a Restaurant Use AI Food Photography on Menus and Delivery Apps? — Where AI food photos are safe (lighting, backgrounds) and where they mislead (portions, ingredients), with a spec-card check and a real-photo fallback.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Photoroom pricing page (October 2026), Amazon Seller Central image guidance as summarised in seller community posts (check the current help page), Adobe Firefly and Canva product documentation.