Complete Ecommerce Product Photography Guide

Start with the product decision, not a moodboard

A product image programme should answer the questions a buyer has before adding an item to cart: what is this exact product, what variant is shown, what does it look like from the required views and what important material or fit details must remain visible? A visual direction is useful, but it cannot replace a product-specific image brief.

For each SKU family, document the hero view, secondary views, detail shots, required crops, background rules, model or flat-lay requirements, file names and destination channels. This gives the production team a stable definition of done and makes later review more objective.

Define the image set before selecting a production method

Treat channel requirements as production constraints

Google Merchant Center requires product images that accurately display the product and sets technical and content requirements for images used in listings. Its guidance also recommends distinct images for visually different variants and calls out product visibility, image quality and stable image URLs. Those rules should be checked during production, not only after a final delivery folder is assembled.

Marketplace, retailer and owned-site requirements can differ. Keep a current channel matrix for aspect ratios, file types, image dimensions, background, crops, naming and metadata. Google Merchant Center also requires generative-AI product images to retain the appropriate IPTC DigitalSourceType metadata; do not remove that metadata during postproduction or delivery. If a frame is made for more than one channel, record which version is the source and which derivatives are approved.

Choose physical, AI-assisted or hybrid production by the unresolved risk

Physical capture is often the best route when texture, movement, fit, reflection or product interaction must be discovered or verified on set. An AI-human production route can fit recurring, bounded work when reliable product inputs, a selected visual reference and clear acceptance rules already exist. A hybrid route can use physical capture for product truth and use controlled AI-assisted work for approved variations or repeatable extensions.

Do not select a route because a demo image looks convincing. Test it with representative product inputs, including the normal difficult cases. The route is ready to scale only when it can preserve the required details, move exceptions to a named owner and deliver the exact file package the channel needs.

Build review into the route, not at the end

The production unit is the approved, usable delivery—not the number of frames captured or generated. Measuring only image volume hides the correction queue, version confusion and client-side effort that make a programme expensive or unreliable.

• Use a golden reference and a written acceptance checklist before the first full batch.
• Check product truth: silhouette, colour, material, branding, print, hardware and variant accuracy.
• Check image-family continuity: camera, crop, lighting, background and model treatment where relevant.
• Separate pass, correct, hold and reject so uncertain frames do not enter delivery unnoticed.
• Record who owns corrections, what is an input problem and when a route should stop for a decision.

Run a representative pilot before scaling

A useful pilot contains normal products and difficult but expected exceptions. Agree the input pack, image set, review window, delivery specification and success criteria in advance. At the end, assess approved outputs, correction load, held inputs and time spent by both teams—not merely the strongest portfolio images.

8M Studio scopes catalogue production around this accountable progression: defined inputs, an agreed route, human QC, exception handling and approved delivery. A paid pilot is the right next step when a team needs to test that production system against its own product and channel requirements.