# Product Photography Workflow: What to Automate With AI (2026) > The product photography workflow in nine steps, shot list to publish: what to keep human, what to automate with AI, and the channel specs that decide it. Published: 2026-10-14 Updated: 2026-10-14 Author: Olga Stogova Tags: product-photography-workflow, product-photography-automation, ecommerce, ai-product-photography, image-specs Kive studio: Urban Respite Canonical: https://kive.ai/learn/product-photography-workflow A product photography workflow is the same nine steps every time: plan, capture, select, retouch, cut out, make variants, export per channel, check, publish. Automate everything after the shutter except the final check, keep capture and selection human, and write down each channel's export rules before you buy a single tool, because that is where automated images get rejected. ## Key takeaways - Keep capture, selection and the final check against the real product human; they are judgments about a physical item. - Background removal costs $0.02 per image on Photoroom's Basic API plan, so a pipeline rarely needs a person for cut-outs. - Etsy shows transparent areas as black, so every transparent PNG from a cut-out step needs a flattened background before upload. - Google Merchant Center requires AI-generated images to carry metadata saying so, which aggressive export settings can strip. - remove.bg's standalone website closes on 1 December 2026, so pipelines built on it need a replacement plan. --- ## The product photography workflow, step by step Nine steps cover a product from arrival to a live listing. The order matters more than the tools: each step's output is the next step's input, so a mistake at capture costs a reshoot, while a mistake at export costs one re-run. ### Nine steps from product to listing **Write the shot list** List every frame per channel before anything is shot: main image, angles, detail, in-scale, lifestyle. Baymard found 42% of users try to judge a product's size from its images, and 37% of sites offer no in-scale image [[1]](https://baymard.com/blog/current-state-ecommerce-product-page-ux), so put a scale shot on the list from the start. **Capture the product** Shoot the real item, straight and evenly lit, with the label sharp. This is the source every later step inherits; no tool recovers a color or a detail the camera never recorded. **Select the frames** Pick one hero per angle and reject soft, reflective or crooked frames. Software can flag blur, but deciding which frame sells the product is a person's call. **Retouch and correct color in batches** Fix one frame, then copy its settings to the rest. In Lightroom, Auto Sync applies settings to multiple images together [[2]](https://helpx.adobe.com/lightroom-classic/desktop/help/copy-paste-settings.html). For hand retouching at volume, Pixelz lists a Professional plan at $0.95 per image with next-morning delivery [[3]](https://www.pixelz.com/pricing/). **Remove or replace the background** Run cut-outs as a batch or through an API. Keep the transparent PNG as a master file and produce flattened copies per channel at export. **Make variants and lifestyle scenes** Generate scenes, colorways or seasonal sets from the approved product photo rather than reshooting. Use a generator that takes your photo as the source, so the product stays the one you sell. **Export per channel** Resize, crop, flatten and name files for each destination in one pass. Capture One can export a variant in multiple file formats, image sizes and color spaces at the same time [[4]](https://support.captureone.com/hc/en-us/articles/360002646058). **Check against the real product** Compare every final image with the item or the source photo: label text, color, shape, logo placement. This is the step automation makes more important, because generated and edited images can drift without looking wrong. **Publish in bulk** Upload through the platform's bulk route. Shopify's Files page takes up to 200 files at once [[5]](https://help.shopify.com/en/manual/shopify-admin/productivity-tools/file-uploads), and a product CSV can carry up to 250 images per product, one row per image [[6]](https://help.shopify.com/en/manual/products/import-export/using-csv). --- ## Which steps to automate, and with what Automate a step when its rule can be written down and checked. Cut-outs, color presets, resizing and uploads pass that test. Capture, selection and the final check don't, which is why they stay human even in fully automated teams. | Step | Who should own it | Automation route | Cost reference | | --- | --- | --- | --- | | Shot list | Person | Reusable template per category | – | | Capture | Person | None for new products | – | | Selection | Person | Blur flagging at most | – | | Retouch and color | Batch | Lightroom Sync or Auto Sync; Pixelz for hand retouching | Pixelz $0.95 per image | | Background | API or batch | Photoroom API, Claid API, editor background tools | Photoroom Basic $0.02 per image | | Variants and scenes | AI generation | Generators that start from the product photo | Kive Basic: 2,000 credits for $40/month | | Export per channel | Batch | Export recipes in Capture One; canvas extend for new ratios | Kive extend 2 credits | | Final check | Person | None | – | | Publish | Batch | Shopify Files or CSV import; platform integrations | – | Photoroom's Basic API plan lists $20 a month at $0.02 per image for background removal, and its Plus plan $100 a month at $0.10 per image for full image editing, billed monthly only [[7]](https://www.photoroom.com/api/pricing). If your pipeline calls remove.bg, note that its standalone website will no longer be available from 1 December 2026 and its background removal is moving to Canva [[8]](https://www.remove.bg/pricing). For the variants step, Kive builds a product once from one to four photos and generates scenes around it with a studio, a saved look of lighting, framing and setting. Its [Bulk generate](https://kive.ai/docs/creating-with-ai/bulk-generate) applies one studio to selected products in a single run, on paid plans and on desktop, so two versions for ten products gives twenty images. It does not batch-edit existing photos, and simple background tools are faster for high-volume cut-outs. [Bulk image generation for product catalogues](/learn/bulk-image-generation-product-catalogue) works through the credit arithmetic for a full run. --- ## Export rules that break automated pipelines An automated image can pass every check in your pipeline and still fail at the destination. Each marketplace sets its own size, background and labeling rules, and an export step that ignores them undoes the time saved upstream. | Channel | Size | Image rules | Watch for | | --- | --- | --- | --- | | Amazon | Zoom from 1,000 px on the longest side | Pure white RGB 255, 255, 255; product fills 85% of the image | File names: product ID + variant code | | Google Merchant Center | 1500 x 1500 recommended; 500 x 500 minimum from 31 Jan 2027 | No promotional elements or overlays | AI images must carry AI-generated metadata | | Shopify | 2048 x 2048 usually best for square | Up to 5000 x 5000 or 25 megapixels, under 20 MB | Up to 250 images per product | | Etsy | At least 2000 px wide and high | First photo at least 635 px | Transparent areas show as black | | Instagram feed | At least 1080 px wide | Aspect ratio between 1.91:1 and 3:4 | Crop per ratio, not per platform | Amazon's main image needs a pure white background with the product filling 85% of the frame, and images of 1,000 pixels or more on the longest side enable zoom [[9]](https://sellercentral.amazon.com/help/hub/reference/external/G1881). Google Merchant Center will require at least 500 x 500 pixels for all products from 31 January 2027 and recommends 1500 x 1500 or above [[10]](https://support.google.com/merchants/answer/6324350). Shopify says a 2048 x 2048 square usually displays best [[11]](https://help.shopify.com/en/manual/products/product-media/product-media-types). Etsy recommends at least 2000 pixels on each side [[12]](https://help.etsy.com/hc/en-us/articles/115015663347), and Instagram accepts widths from 1080 pixels at ratios between 1.91:1 and 3:4 [[13]](https://help.instagram.com/1631821640426723/). Size guides per platform are collected in [Etsy listing photo size](/learn/etsy-listing-photo-size). **Two rules automation breaks quietly** Etsy renders transparent parts of an image as black [[12]](https://help.etsy.com/hc/en-us/articles/115015663347), so a cut-out PNG uploaded as-is ships with a black background. And Google Merchant Center requires every image created with generative AI to contain metadata saying so, such as the IPTC DigitalSourceType value TrainedAlgorithmicMedia [[10]](https://support.google.com/merchants/answer/6324350). Check that your compression and resize settings keep it. --- ## GUI, API or agent: picking the automation level Pick the automation level by who runs the step, not by volume alone. A GUI tool keeps a person clicking, an API removes the person from a fixed step, and an agent lets a person describe a job in a sentence and review what comes back. | | GUI tool | API | AI agent over MCP | | --- | --- | --- | --- | | Who runs it | A designer or seller | A developer, once | Anyone who can describe the job | | Best for | Tens to hundreds of images, varied edits | One fixed step at any volume | Mixed jobs that change week to week | | Setup | None | Code, keys, error handling | Connect the tool to Claude, ChatGPT or another client | | Weak spot | Repetition at scale | Rebuilding when the step changes | One request per product; review every result | | Example | Lightroom presets, editor background tools | Photoroom or Claid API | Kive, Photoroom or Adobe connectors | _Examples from each vendor's own documentation, checked October 2026._ Agents are the newest route. Kive's MCP server works with ChatGPT, Claude, Cursor, Figma, Notion, Zapier and other clients, and lets them create products, generate images and edit them from a sentence ([Kive MCP overview](https://kive.ai/docs/mcp/overview)). It has no batch tool and Kive has no public REST API for generation, so a catalog-wide run still belongs in Bulk generate. What an assistant can and cannot do with images natively is covered in [can Claude generate images](/learn/can-claude-generate-images). --- ## Where shoppers look first Images are where most shoppers start on a product page, so the last check deserves the most attention. In Baymard's usability testing, 56% of users' first action on a product page was to start exploring the product images, and 25% of e-commerce sites provided images insufficient for a close inspection [[14]](https://baymard.com/blog/ensure-sufficient-image-resolution-and-zoom). ### What that means for the checklist Two of the steps above carry most of that weight. The shot list decides whether a shopper can judge size and use, and the export step decides whether they can zoom. Both are cheap to get right and expensive to retrofit across a live catalog, which is why they belong in the written workflow rather than in someone's memory. For what to look for when the image is generated rather than shot, see [how to make AI product photos look real](/learn/make-ai-product-photos-look-real). ### Automate the middle, keep the ends Capture and the final check stay human. Retouching, cut-outs, variants, export and upload are rules, so let software run them. - **Solo sellers editing by hand:** A GUI editor with presets – batch color and background without code - **Stores with a developer:** Photoroom API – $0.02 per cut-out on the Basic plan - **High-volume hand retouching:** Pixelz – $0.95 per image, next-morning delivery - **Lifestyle variants across a catalog:** Kive Bulk generate – one studio applied to many products in one run Write the workflow as a one-page document with an owner per step and the export rules for every channel you sell on. That page is what lets you swap a tool later without breaking the pipeline, and the remove.bg change in December is a reminder that the tools will change. Start automation at the step where your team spends the most hours per week, measure it for a month, then move to the next. Keep the final check human throughout. ## Sources 1. [Baymard Institute – Product Page UX: current state]() 2. [Adobe Help – Copy and paste edit settings in Lightroom]() 3. [Pixelz – Pricing]() 4. [Capture One Support – Processing multiple variants]() 5. [Shopify Help – Uploading files]() 6. [Shopify Help – Using CSV files]() 7. [Photoroom – API pricing]() 8. [remove.bg – Pricing]() 9. [Amazon Seller Central – Product image requirements]() 10. [Google Merchant Center Help – Image requirements]() 11. [Shopify Help – Product media types]() 12. [Etsy Help – Listing photo requirements]() 13. [Instagram Help – Photo resolution]() 14. [Baymard Institute – Sufficient image resolution and zoom]() ## FAQ ### What is a product photography workflow? It is the fixed sequence every product image goes through: shot list, capture, selection, retouching, background, variants, export per channel, quality check and publishing. Writing it down lets you see which steps a person must own and which can run as a batch, an API call or an agent task. ### Which product photography steps can AI automate? Background removal, batch color correction, lifestyle variants from an existing product photo, resizing and export per channel, and bulk upload. Capture of a new product, choosing the hero frame and checking the result against the real item should stay with a person, because each is a judgment about the physical product. ### What image size works for Amazon, Shopify and Etsy? Amazon enables zoom at 1,000 pixels or more on the longest side and wants the product to fill 85% of a pure white main image. Shopify says 2048 x 2048 usually displays best for square images. Etsy recommends at least 2000 pixels on width and height. ### Do AI-generated product images need to be labeled? On Google Merchant Center, yes. Google requires every image created with generative AI to carry metadata saying so, for example the IPTC DigitalSourceType value TrainedAlgorithmicMedia. Check that your export step keeps that metadata instead of stripping it while resizing or compressing. ### Should I use a GUI tool, an API or an AI agent? A GUI tool suits a person working through tens or hundreds of images by hand. An API suits a developer wiring a fixed step, such as cut-outs, into a store or DAM. An agent suits a person who wants to describe the job in a sentence and review the results, without writing code. ### Can Kive automate the lifestyle-image step? Yes, for variants. Bulk generate applies one Studio to selected products in a single run, on paid plans and on desktop, so two versions for ten products gives twenty images. It does not batch-edit existing photos, and new products still need a real photo first.