White-background images for 200 SKUs: a batch workflow from shoot to AI cutout and QA

Sellers of small goods, hardware and household items know that launching dozens or hundreds of SKUs at once is normal. Every SKU needs at least one white-background image, some platforms want several angles, and cutting them out one by one in an editor means even a fast designer manages a few dozen a day — fewer with tricky edges.

Using a batch of 200 kitchen tools as the example, this is the workflow we actually use. The idea: prepare for cutouts during the shoot, let AI do the cutting, and spend human effort on review and difficult products.

Step 1: shoot for the cutout

Cutout quality is capped at the shoot:

• Contrast product and background — mid-grey for light products, light backgrounds for dark ones. Avoid shooting white products on white paper; the edges disappear.
• Even light — two soft lights from either side to reduce hard shadows. Heavy shadows get mistaken for part of the product.
• Fixed camera — same position and focal length for the whole batch, so products appear at the same size and angle.
• One product per frame with room around it.
• Sharp focus — blurry edges are hard to cut cleanly, AI or not.

Step 2: group by difficulty

Don't throw all 200 in at once. Group by how hard they are to cut out:

In our kitchen batch, group A (silicone spatulas, wooden spoons, measuring cups) was about 60%, group B (whisks, colanders, graters) about 30%, and group C (glass spice jars, mirror-finish lids) about 10%.

Step 3: batch AI cutout

AI cutout accepts several images per batch; the per-batch limit depends on your account settings. Key settings:

Output background

• White — pure white, ready to upload as a white-background main image;
• Transparent — a transparent PNG for compositing scenes or placing on different colours later;
• Custom colour — a set background colour for shops that use one light tone throughout.

Our advice: output transparent versions first as masters, then produce white versions as needed. A transparent master can be reused for scenes and colour changes without cutting again.

Edge feathering

• None — sharpest edges, for hard products with clean outlines;
• Soft — slightly softened, right for most products; our usual setting;
• Smooth — for fluffy or fabric products whose edges are soft anyway.

Too much feathering leaves a grey halo on white, one of the common reasons platforms reject white-background images.

Step 4: review

Review is the most important step and the one most often skipped.

Group A: spot-check

Randomly pick two or three of every twenty and check edges at 100%. If they all pass, the group passes; if not, widen the check.

Group B: check every edge

Openings and thin wires are where AI cutouts go wrong: background trapped between whisk wires, colander holes filled in. Check each image zoomed in and fix problems with Inpaint or reprocess.

Group C: handle individually

Transparent and reflective products often come out with glass turned solid white or reflective areas cut away:

• Transparent products: keep the transparent version, and when compositing onto white make sure the glass shows white through it, not the original background colour;
• Reflective products: if reflections show the studio, use reflection removal in Retouch;
• Fluffy products: use smooth feathering and check that clumps of fibre haven't been clipped.

Step 5: naming and archiving

Two hundred SKUs with at least two versions each needs order. Our naming pattern:

SKU_colour_angle_version.png, e.g. K1032_red_45_alpha.png and K1032_red_45_white.jpg

Archive in three folders: originals, transparent masters, white finals. Upload only from the white finals folder.

Step 6: final check before upload

1. Background pure white (sample all four corners);
2. Product complete, not cropped;
3. Consistent product size across images;
4. Clean edges, no halo or leftover background;
5. File names match SKUs.

Platform requirements vary; see Taobao's five main images and product images for JD.

Where the time goes

For these 200 kitchen tools, our time was roughly: one day shooting, half a day batch cutting and grouping, one day reviewing and fixing (mostly groups B and C), and half a day naming and archiving. Compared with cutting everything by hand, the saving is mostly in group A; groups B and C still need people. That's why grouping is the key step — it decides where human attention goes.