Cut-outs, edges and shadows
Most tools handle a matte box on a plain background. The differences show on hard products: transparent and translucent items such as glass bottles and perfume, reflective metal such as jewellery and cookware, fine edges such as fur, hair and knitwear fibres, and white products on white. Ask for output with an alpha channel or clipping path kept in the master file, so the mask can be reused. Shadows need the same scrutiny: a natural contact shadow grounds a product, a generic drop shadow looks pasted on, and a catalogue needs one shadow direction and softness throughout. Trial with your worst products, not the vendor's samples.
Retouching that leaves the product alone
Removing dust, scratches, fingerprints and creases in packaging, straightening labels, taking out mannequins and props: all normal. Changing the product itself is not. Altering shape, texture, stitching, features or the apparent effect of a cosmetic can mislead buyers, and UK advertising guidance on cosmetics warns specifically against retouching that exaggerates how a product performs. Generative fill is the risk to watch: asked to extend a background or remove a reflection, it can quietly invent detail at the product's edges. Write down what edits may and may not change, keep the original capture, and log each edit against the file.
Colour accuracy is measurable
In fashion, beauty and home, colour errors become returns, because customers send back what arrives looking different. Treat colour as data. Shoot with a colour reference chart, keep ICC profiles embedded through every step, convert consistently to sRGB for the web and measure drift against a reference swatch using Delta E. Automatic enhancement and generative relighting often push saturation or shift hue, so switch them off for colour-critical categories. When a tool recolours one product photo into another colourway, check the result against the physical swatch, not against the colour name in the PIM.
Batch jobs with marketplace checks
A catalogue edit runs across thousands of files, so consistency and checks matter more than any single result. Look for saved settings applied identically across a batch; automated checks against each destination's rules, such as a pure white background for Amazon main images, minimum dimensions and no added text or watermarks; and an exception queue that sends failures to a person. Synthetic White fits where editing meets generation: batch jobs run from a PIM or product feed, product-lock keeps each product's geometry, labels and colours pixel-accurate while backgrounds and scenes change, and finished files go to the DAM through connectors.
Updated 25 September 2026