There are four kinds of retouching supplier, they solve four different problems, and the expensive mistake is almost always buying the wrong kind rather than the wrong company.
This page sets out what each kind is genuinely good at and where each one fails. That includes where we fail, which is at the bottom and is the part worth reading if you are comparing us against anyone.
Best for: a simple, repeatable operation defined in advance, at large volume, inside your own pipeline. Background removal, format and crop conversion, marketplace compliance. Per-image cost is low, turnaround is effectively instant, and it sits behind an API so your PIM or storefront can call it directly.
Wrong for: anything where the correct result depends on knowing what the physical product looks like. A model cannot check a colour against a garment it has never seen, and it cannot know which of two plausible neck shapes is the one that was manufactured. It produces something confident and wrong, at volume, quickly.
Best for: clipping path and background removal on simple silhouettes, where the cheapest acceptable result is the right commercial answer. If the brief is "cut it out cleanly and put it on white" and the shape is uncomplicated, paying more buys you very little.
Wrong for: anything that has to match across a range. The failure mode is not bad images, it is inconsistent ones — image 1 and image 4,000 handled by different people to different unwritten assumptions. You will not see it in a sample of five. You will see it on the product listing page.
This is where we sit, so read the next section as well as this one.
Best for: thousands of SKUs that must look like they came from one place. The thing you are buying is not retouching, it is a written standard and the QC that enforces it. Ghost mannequin, apparel, packshot, flat lay and accessories at catalogue volume, held to one specification across shifts and seasons.
Wrong for: small volumes and single images. See below.
Best for: a handful of images where every pixel is looked at — campaign, editorial, beauty, the hero shot that carries a season. One artist, deep judgement, a real creative conversation.
Wrong for: volume and reliability. One person has one pair of hands, gets ill, and takes holidays. A turnaround promise from a single retoucher is a promise about one person's calendar.
Most supplier comparisons turn into price lists. These seven questions separate suppliers faster than a rate card, and every one of them has a checkable answer:
Buyers usually arrive here having already found one large, well-known high-volume service and wanting to know what else exists. The useful answer is not another company name — it is the question of which of the four kinds above your problem actually is.
If your volume is large and the operation is simple and repeatable, an AI platform will beat every human studio on cost and speed, including us. If your volume is large and the work needs judgement about a physical product, you are choosing between managed studios, and then the differences are the seven questions above. If your volume is small, you should not be talking to any of us.
This is the part most supplier pages leave out, so it is the part worth trusting.
A white-label retouching studio in Dhaka. 58 in-house retouchers and 6 QC managers across three shifts, three project managers, 40,000+ images a month, six days a week. Free test edit back in six hours; live work in 24–48 hours. Rates published on the site rather than quoted on request. We sign the client's DPA, and client files are never used to train any model.
Every figure in that paragraph is checkable on this site without asking us, which is the only reason to put it in writing.
Four kinds of supplier: AI platforms and APIs, low-cost bulk vendors, managed high-volume studios, and boutique or individual retouchers. They solve different problems, and the expensive mistake is usually buying the wrong kind rather than the wrong company within a kind.
If the images are unreleased product, yes — and ask for a Data Processing Agreement as well. An NDA covers confidentiality; a DPA covers how your files are processed, where they are stored, who can access them and when they are deleted. Ask whether the studio's own subcontractors are bound by equivalent terms, because many high-volume suppliers use them.
When the operation is simple, repeatable and defined in advance, the volume is large, and you want it inside your own pipeline through an API. AI is strong on background removal and format conversion at scale, and weak wherever the correct result depends on knowing what the physical product looks like.
Below roughly 500 images a month, for a single hero image, or for creative art direction. In all three cases a high-volume studio is the wrong tool — including ours.
Send five of your own images, and send the difficult ones rather than the clean ones. Anyone can do an easy frame. Then ask them to explain a decision they made on one of them; the answer tells you whether a person looked at it.
Send the difficult ones. Anyone can do an easy frame.