Background Removal: DIY, AI and a Retoucher Compared
A clean background is the cheapest click-through improvement in product photography, and there are three ways to get one: shoot it white, remove it with AI, or pay someone. Each wins in a different situation. This guide compares cost, time, quality and failure modes, then shows how to fix the edges that give AI cut-outs away.
The three options, honestly described
Shooting on white is the slowest per photo and the most natural result, especially for reflective, transparent or fuzzy products. AI removal is the fastest and works well on hard-edged objects, but it guesses at ambiguous edges. A retoucher costs money per image and handles the cases the other two cannot.
Most catalogues end up with a hybrid: shoot on a clean background when possible, AI-remove when the product is simple, and pay for the handful of hero images that carry the brand.
Comparison at a glance
Treat the time column as typical, not guaranteed. The first photo of a session always takes longer than the twentieth, and a badly lit original makes every method slower.
| Method | Typical time per image | Cost | Best for | Main weakness |
|---|---|---|---|---|
| Shoot on white | 5–15 min incl. setup | Low (materials) | Glass, jewellery, hair, texture | Consistency across a batch |
| AI removal | Seconds + review | Free to low | Hard-edged products, bulk | Hair, chains, translucent edges |
| Retoucher | Hours to days turnaround | Per image | Hero shots, complex products | Cost and feedback loop |
Shooting on white without a studio
You need three things: diffuse daylight, a large white surface, and separation from the background. A window on an overcast day is a better light source than most cheap softboxes because it is large and soft.
Tape a sheet of white paper or foam board to a wall and let it curve onto the table — a sweep, not a corner. Put the product a little forward from the wall so it does not cast a hard shadow on the background. Shoot at eye level or slightly above, fill the frame, and lock white balance instead of relying on auto.
If the product is white, add a grey card next to it for the first frame and remove it after setting exposure. A white product on a white sweep without separation is the most common cause of a “flat” main image.
Where AI removal wins — and where it fails
AI segmentation is excellent at boxes, bottles, shoes, furniture and anything with a clear silhouette. It is fast enough that reviewing ten outputs takes less time than masking one by hand.
It struggles with four categories: hair and fur, transparent or reflective materials, thin straps and chains, and products that are the same colour as the background. In those cases the model invents an edge, and the mistake is visible at 100% even if it looks fine in a thumbnail.
- Wins: boxed goods, ceramics, bags, hard toys, books, packaged food.
- Fails: glasses, perfume bottles, jewellery chains, fluffy textiles, white-on-white.
- Sometimes: dark products on dark backgrounds, or products with soft shadows you want to keep.
Hiring a retoucher without wasting money
Send a brief, not a folder. The brief should say: output size and format, background colour (pure white or transparent), whether shadows should be kept, and whether the product must remain pixel-accurate. Ambiguity is what turns a one-hour job into three revisions.
Ask for one test image before booking a batch. A good retoucher will return a cut-out with a clean edge and a believable contact shadow; a weak one will return a hard mask with a white halo. The test costs little and tells you everything.
Fixing the edges AI leaves behind
Three artefacts give an AI cut-out away: a bright halo along the edge, a missing contact shadow, and a jagged silhouette on curves. Each has a quick fix.
- Halo: contract the mask by 1–2 px, or paint a slightly darker edge on the product side.
- Missing shadow: add a soft ellipse under the product at low opacity, blurred, so it appears grounded.
- Jagged curve: feather the mask 0.5–1 px; a hard edge reads as fake at large sizes.
- Colour fringe: desaturate the edge 10–20% or clone the product colour inward.
- Floating object: check the baseline — the lowest point of the product should sit on the same line across the gallery.
White background: required or not?
Most marketplaces do not require a pure white background, but the highest-converting main images are usually clean, bright and high-contrast. Some categories and some platforms have stricter rules — always check the current image policy for the specific category before reshooting a catalogue.
For your own store, consistency beats purity: the same background tone across a collection makes the grid look curated, even if it is a warm off-white rather than #FFFFFF.
Choose by catalogue size
| Catalogue size | Recommended approach |
|---|---|
| Under 20 products | Shoot on white; reshoot rather than mask |
| 20–100 products | Shoot on white + AI for simple shapes |
| 100–1000 products | AI first pass, human QA sample, retoucher for heroes |
| 1000+ products | Standardised capture setup + AI + automated QA checks |
The thumbnail test that decides everything
Shrink the finished main image to about 200 px wide and look at it on a phone. If you cannot tell what the product is, the background work failed regardless of how clean the edge looks at full size. Fix the crop and contrast before you fix the mask.
Frequently asked questions
Is AI background removal good enough for Etsy?
For hard-edged products, usually yes. Check the edges at 100% and add a contact shadow so the item does not look like it is floating.
Should the main image have a pure white background?
A clean, bright background wins clicks. Pure white is not universally required; check the policy for your category and keep the collection consistent.
Why does my cut-out have a white halo?
The mask included light pixels from the old background. Contract the mask slightly and darken the edge on the product side.
Can I remove backgrounds from 50 images at once?
Yes with a batch AI tool. Review a sample of five and inspect complex products individually.
Do I lose quality by removing the background?
Only if you re-encode repeatedly. Keep a lossless or high-quality master and export the final JPG or WEBP once.
About BulkImageSEO
BulkImageSEO is a free image SEO toolkit maintained by Salah Berrehou. It renames files with keyword-rich names, injects Title, Subject, Tags and Comments metadata, compresses images for Core Web Vitals and converts formats. Uploaded files are deleted after 60 minutes and no account is required. Read more about the author, the step-by-step guide, or contact us.