# Remove background from photo: AI product photos for online stores Remove the background from a product photo with AI: 8 tools compared, the best background, phone photos in 8 steps and Google and Meta image sizes. Source: https://cloudcart.com/en/blog/ai-product-photos · CloudCart · 2026-11-06 In short: you can remove the background from a product photo in seconds with an AI editor such as remove.bg, Adobe Firefly, Remove Background in Word and PowerPoint, Product Studio in Google Merchant Center, ChatGPT, Gemini or your phone's own Photos app. For the main image, use a white background, let the product fill 75-90% of the frame and export at least 1500 x 1500 pixels, as Google recommends. Below: eight ways to remove a background, the right background for each photo, phone photography in 8 steps, the sizes Google and Meta expect, and a full catalog check in CloudCart. It applies to sellers anywhere in the EU. The CloudCart facts come from the platform documentation and the [Catalog](https://cloudcart.com/en/catalog) and [AURA](https://cloudcart.com/en/aura) pages. ## How to remove the background from a product photo: 8 ways An AI background remover detects the product and deletes everything around it, leaving a transparent PNG or a photo on white. A web tool is enough for 10 photos; 500 photos need batch processing. | Tool | How it works | Best for | Watch out for | | remove.bg and iLoveIMG | Upload, the background goes in seconds, download PNG or white | Single photos, fast | Full resolution and batches are on paid plans | | Adobe Firefly and Adobe Express | Upload, download a transparent PNG, add a new background in the same editor | Sellers already using Adobe | Edges of glass and glossy products | | Your phone (iPhone, Android) | iOS 16 or later: in Photos, touch and hold the product, then Copy or Share; many Android galleries can cut out objects too | Shoot and edit on the spot | You get a transparent PNG; white is a separate step | | Adobe Photoshop and GIMP | Photoshop: Properties → Quick Actions → Remove Background; GIMP: manual selection | Hair, fur, glass, jewellery | Needs time and skill | | Word and PowerPoint (Microsoft 365) | Picture Format → Remove Background, then Mark Areas to Keep or Remove, then Keep Changes and Save as Picture (PNG) | A few photos, if you have Office | Fine edges often need correction | | Product Studio in Google Merchant Center | Removes and replaces backgrounds, upscales, creates scenes | Stores selling on Google Shopping | Keep the AI metadata it adds | | ChatGPT and Gemini | Upload and write "remove the background and replace it with pure white" | Adding a shadow, scene or colour too | May alter a label, seam or shade | | A script run by an AI agent | Claude Code or Codex runs the open-source rembg over a whole folder | Hundreds of photos, identical framing | Review the first 10 results first | Rule of thumb: for 1-10 photos use your phone, remove.bg or PowerPoint; for 10-100 use remove.bg or Adobe Firefly and check the edges; for 100 or more let an AI agent run rembg over the folder and upload by product code (SKU). Glass, fur and jewellery need an AI tool plus 5-15 minutes of manual correction per photo, by our estimate. Important: any AI tool can change the product, not only the background. Compare the edited photo with the original at 100% zoom: colour, label, handles, transparent parts. In the EU a shopper can withdraw from an online purchase within 14 days of delivery without giving a reason ([Your Europe](https://europa.eu/youreurope/citizens/consumers/shopping/returns/index_en.htm)), so a product that arrives in a different shade than the photo usually comes straight back. ### How to put a product on a white background or change the background A white background takes two steps: first remove the background to get a transparent PNG, then place the product on a pure white square (#FFFFFF). In remove.bg and Adobe Firefly white is one click, and in ChatGPT and Gemini you write it in the instruction. Changing the background to another colour or a scene works the same way: you pick the new background instead of the white square. ## What is the best background for product photos? For the main product image, the best background is white or very light and plain, so the product looks consistent on category pages, in Google Shopping and in the Meta catalog. Extra photos can show a setting. | Photo | Background | Why | | Main image (first in the gallery) | White or light grey, no props | A consistent category page; Image Health in CloudCart flags a busy background on the main image for Google | | 2-3 photos from other angles | The same white background | The shopper compares details without distraction | | Scale photo | A setting or a hand next to the product | In Baymard testing, 42% of users tried to gauge a product's size from its images | | In-use (lifestyle) photo | A real setting | Shows who the product is for; Google accepts it as a separate image in the lifestyle_image_link attribute | Google sets clear rules for the main image in Merchant Center. The product should fill roughly 75% to 90% of the frame. The image must have no promotional text, price, "Buy now", watermark, added store logo, border or collage of several products ([Google Merchant Center, image_link requirements](https://support.google.com/merchants/answer/6324350?hl=en)). Images that break these rules show up under Diagnostics in Merchant Center and are not shown. How to get the catalog into Google is covered in the guide to [Google Merchant Center and Google Shopping](https://cloudcart.com/en/blog/google-shopping-merchant-center-bulgaria). ## How to take product photos with your phone: 8 steps A phone gives store-ready product photos when the light, background and distance stay the same for every product. All you add is white card and a small tripod. - Window light. Put the table next to a large window, out of direct sun. Switch off the room lights so you don't mix two colour temperatures. - A white background with no corner. Curve a sheet of white card from the wall down onto the table, like a sweep. There is then no visible line between wall and table. - A reflector. A second white card opposite the window lights the shadow side. - Tripod and timer. The tripod keeps framing identical; a 3-second timer stops shake. - The main lens, not the ultra-wide. The ultra-wide lens distorts edges. Step back and use 2x if your phone has it. - Lock focus and exposure. Hold your finger on the product until they lock, and lower brightness slightly if the white bleeds into the product. - The same set of shots. For every product: front, 45° angle, back, a detail and a scale photo. Image Health in CloudCart expects at least 3 photos per product by default. - Shoot at full resolution. Turn off filters and portrait mode. Remove the background afterwards with a tool from the table above. Example: in a [handmade ceramics store](https://cloudcart.com/en/blog/handmade-online-store), this setup gets 40 products photographed in one afternoon, with backgrounds removed for the whole folder at once and the same light on every photo. > Store for €1 36 photos, a video and a 360° view per product, at €1/month for the first 3 months The gallery of up to 36 photos, the YouTube video and the 360° image of up to 36 frames are in the Media box of each product . New stores until 31.12.2026 pay €1/month for the first 3 months on every plan, with 0% commission on your sales . Create a store for €1 → ## What size and format should product photos be? Product photos for an online store should be at least 1500 x 1500 pixels, square, in JPG or WebP. From 31 January 2027 Google will require at least 500 x 500 pixels for every product and recommends 1500 x 1500 or larger ([Google](https://support.google.com/merchants/answer/6324350?hl=en)). A square image displays uncropped on your category pages, in Google Shopping and in the Meta catalog. | Requirement | Value | Source | | Google minimum, from 31.01.2027 | 500 x 500 px | Google Merchant Center | | Google recommendation | 1500 x 1500 px or larger | Google Merchant Center | | Google maximum | 64 megapixels and 16 MB per file | Google Merchant Center | | Below this, Image Health flags the photo as too small for the Meta catalog | 500 px on the short side | Image Health in CloudCart | | Below this, Image Health flags low resolution for zoom | 800 px on the short side | Image Health in CloudCart | | Unusual proportions | The long side is more than 3 times the short side | Image Health in CloudCart | | Formats CloudCart accepts | JPG, PNG, WebP, GIF, SVG (for SVG and GIF, Image Health offers conversion to JPEG) | The product in the admin | | Photos per product in CloudCart | Up to 36, plus one YouTube video and one 360° image of up to 36 frames | The product in the admin | CloudCart creates the smaller storefront sizes itself and serves them from a CDN, so you upload one full-size file. The first photo in the gallery is the main image; drag to reorder or use Set as primary. How photos affect orders placed on phones is covered in the article on the [mobile version of an online store](https://cloudcart.com/en/blog/mobile-version-online-store). ## Can AI create product photos? Yes, AI can create product photos, and it works best starting from a real photo: it removes the background, adds a shadow, upscales and places the product in a scene. Keep the main image a photo of the actual product, because the shopper compares the parcel with it, and Google requires the image to show exactly the item being sold. Use AI for scenes on extra photos (a mug on a kitchen counter, useful for Meta and TikTok ads, see the guide to [TikTok Shop and TikTok ads](https://cloudcart.com/en/blog/tiktok-shop-tiktok-ads)), for background variations and for cleaning up dust and reflections. Don't recolour a product to "create" a variant you never photographed: Google wants a separate image for each variant in its exact colour, pattern and material. Google requires every image made with generative AI to carry metadata saying so, such as the IPTC DigitalSourceType tag with the value TrainedAlgorithmicMedia, and forbids removing it. Tools such as Product Studio add it automatically. The EU AI Act points the same way: its transparency rules apply from August 2026, and providers of generative AI must make the content they produce identifiable ([European Commission](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai)). In practice, keep the marking your tool adds and don't run AI images through software that strips metadata. In CloudCart, Aura Studio can read a product photo you select in the description panel (with VisionSense AI on) and put the colour and material into the text, for example "matte green finish and black lid", so photo and description match. > AURA Aura Studio writes the description from the photo and the product data Once the photos are ready, Aura Studio writes the description, SEO title and URL for whole lists of products at once. Every text opens in the editor and goes into the product only when you accept and save it. What AURA does → ## How to check every photo in your store in CloudCart In CloudCart, the Image Health app checks every photo in the catalog and gives the store a score out of 100, starting with the main images of the best sellers. It checks size, format and photo count, and uses AI to spot blur, watermarks, text, busy backgrounds, cropped products and photos that don't match the product. - Install the app. Apps → Image Health. It installs on any plan and has no on/off switch: installed means running. - Set your channels. On the Settings tab, in the Checks box, choose where you sell (Google, Meta) and how many photos each product should have. The default is 3. - Run a scan. On the Health score tab, click Scan catalog. The scan runs in the background and takes a few minutes for a large store. - Open Issues. At the top are the products Google or Meta reject, then the best sellers. The "Revenue, 90 days" column shows 90-day revenue for each product with a problem. - Fix from the product panel. Apply the fix, dismiss the problem or upload a new photo via Edit product. - Keep the automatic check on. With "Check new and changed images automatically", every new or replaced photo is checked about 10 minutes after the change. | Fix in Image Health | When it is offered | Credits | | AI alt text for one photo or the whole catalog | A photo without alt text | 1 per photo | | Link a photo to a colour | A variant colour with no photo | 1 per photo, up to 10 | | A better main image | An AI problem on the main image, when there is more than one photo | 1 per photo, up to 3 | | Square with white margins | Unusual proportions | 0 | | Convert to JPEG | Unsupported format or GIF | 0 | Every store gets 50 credits a month, one credit per photo checked, and monthly packs cover larger catalogs. The score stays visible even when credits run out. AI suggestions change nothing until you apply them, and white margins and format conversion keep the original behind "Restore original". How AI helps with the store's other jobs, from chat to promotions, is in the article [AI for ecommerce: 12 tasks AI can do in your online store](https://cloudcart.com/en/blog/ai-online-store). ## How an AI agent sorts out the photos of a whole catalog An AI agent such as Claude Code, ChatGPT Codex, Gemini CLI or Cursor finds products without photos, processes the photo folder and uploads each photo to the right product in one go. It connects to the store with a PAT token (a personal access token) that you create in the CloudCart admin, and works through the GraphQL Admin API with only the permissions you give it. - Create a token. The store owner opens Settings → PAT tokens and chooses Custom Permissions: Orders: Read for sales data and Products: Read & Write to check and upload photos. Setting up the token is explained in the article on the PAT token in CloudCart . - Ask for a list. "Make a table of products with no photo or fewer than 3 photos, sorted by sales over the last 90 days." - Process the photos. "The Photos folder has one subfolder per product code. Remove the background, put each product on white, make them square at 1500 x 1500 and save them in a folder called Ready." - Review the result. Open 10 finished photos and compare them with the originals before the agent uploads the rest. - Upload them. "Upload the photos from the Ready folder to the products by their code, with the first one as the main image." The agent adds the photos to each product through the CloudCart API. - Measure the effect. Run a new Image Health scan, and after 30 days ask: "Compare sales of the products with new photos, 30 days before and after the change." Orders: Read from step one covers this report. The difference is time. For 300 products, doing it by hand means 300 uploads, 300 cut-outs and opening 300 products in the admin, which by our estimate is several working days. The agent works by product code through the API in an afternoon, and you only review 10 photos and approve. Which numbers to watch after changing the photos is in the article on [ecommerce metrics with an AI agent](https://cloudcart.com/en/blog/ecommerce-kpi-metrics-ai-agent). Why the agent can take over the rest of the store's routine too is explained in the article on [why to run your store with an AI agent](https://cloudcart.com/en/blog/ai-agent-online-store-why). Bulk changes to prices and texts are in the article on [changing the whole catalog with one sentence](https://cloudcart.com/en/blog/bulk-catalog-changes-ai). > Store run by an AI assistant Photos for the whole catalog from one sentence, no manual uploads Claude Code, ChatGPT Codex, Gemini CLI and Cursor connect to your CloudCart store with a PAT token. The agent finds products without photos, processes them on your computer and uploads them, and you grant only the permissions you want. How the AI assistant runs your store → ## Product photo mistakes that cost you orders The costliest photo mistakes get products rejected by Google and Meta or bring goods back as returns. Image Health shows each one within minutes. - Text and discounts on the main image. "-20%" or "New" on the photo gets it rejected by Google. Put labels in the store, not in the image. - One photo for every colour. The shopper picks the blue variant and sees a red one. Image Health flags a colour without a photo as a high-severity problem. - Small photos. A photo under 800 pixels doesn't zoom clearly, and under 500 Image Health flags it as too small for the Meta catalog. - No scale photo. The shopper can't tell how big the bag is, so they ask in the chat or don't order. - Empty alt text. Alt text tells Google what is in the image ( Google Search Central ). In CloudCart, the image name is its alt text. Good photos raise conversion together with the description, the price and the cart. Which other changes bring the most orders is in the article [Online store conversion rate: formula, benchmarks and 20 changes](https://cloudcart.com/en/blog/online-store-conversion-rate), and photos for ads are in the guide to [Meta Ads for an online store](https://cloudcart.com/en/blog/meta-ads-online-store). Questions about photos in your store: support@cloudcart.com. ## Frequently asked questions **How do I remove the background from a product photo?** Upload the photo to an AI background remover such as remove.bg, Product Studio in Google Merchant Center, ChatGPT or Gemini, or use Remove Background in Word and PowerPoint. On an iPhone, touch and hold the product in Photos and copy it without the background. The tool cuts out the product in seconds, and you download a transparent PNG or a photo on white. Compare the result with the original before uploading it to your store. **Can AI create product photos?** AI can process a real photo: remove the background, increase the resolution and place the product in a scene. The main image is best kept as a photo of the actual product, because the shopper compares the parcel with it. Google requires images made with generative AI to carry metadata saying so. **How do I check that my store's photos meet Google and Meta requirements?** In CloudCart, install Image Health from Apps and run Scan catalog. The app gives a score out of 100, shows the photos Google or Meta reject and offers fixes such as AI alt text, white margins and a better main image. Every store gets 50 credits a month, one per photo checked. **Can ChatGPT or Claude process all the photos in my store?** Yes, as an AI agent connected to the store. Claude Code or ChatGPT Codex with a CloudCart PAT token finds products without photos, processes the photo folder on your computer and uploads the photos to the products by their code. Review the first 10 finished photos before the agent uploads the rest.