Pixelz turned Photoshop into an assembly line.

At Trove, the resale business, a good photograph has to do something slightly awkward: show you what is wrong with the clothes. A mark on a used jacket can be information a buyer needs. Remove it, and a technically cleaner image becomes a less useful one. In Pixelz’s account of the partnership, studio manager Lindsey Di Ruscio described the problem neatly: “We’re photographing flaws. This goes against the retoucher’s instincts.”

Trove’s previous supplier had struggled to deliver images before nine in the morning. That meant waiting, chasing, and difficulty tracking progress. Pixelz took on background clipping, hanger and pin removal, and aspect ratios, with instructions to preserve the evidence of wear. The revealing detail is the brief. Before a system can improve an image, someone must decide what improvement means.

Borrow a factory. Edit a photograph.

Pixelz began in Holstebro, Denmark, in 2011, under the wonderfully literal name Remove the Background. Its founders had experienced post-production bottlenecks while running an e-commerce business. Their eventual response borrowed from lean manufacturing: divide an edit into smaller operations, then route each operation to the right resource. The company began rolling out its Specialist-Assisted Workflow, or S.A.W., in 2016.

The early diagnosis was refreshingly specific. Planning left some editors overloaded while others waited. Experienced retouchers spent time on simple cut-outs. Quality assurance became a queue; rejected work made another trip through production. Pixelz’s custom Photoshop software recorded editing activity, giving the company a way to see where time disappeared.

Today, an operation might go to a human specialist, an AI model, or a script. Pixelz says more than half its workflow is automated using AI. That describes a share of the process, rather than a promise that half the people disappear. The useful unit is the task.

The scale accumulated over years. In February 2022, Pixelz reported that it had processed more than 45 million images. That history provides both production experience and a record of recurring requirements. The company can study the routine details that make a catalog coherent: backgrounds, margins, alignment, and color.

Keep the person who knows the hair

Nuuly, the clothing-rental service in the URBN group, provides a practical picture of how to use that capacity. When its studio started, creative director Alison Gates had eight people. Rather than guess how many retouchers future growth would require, she paired Pixelz with an internal retouching specialist who understood the brand and managed quality.

There was trial and error. Messy hair, for example, might be intentional. An external editor could tidy away the look the art director wanted. Nuuly worked out which repeatable operations belonged outside and which aesthetic judgments belonged inside. Outsourcing required someone who could recognize the difference.

One minute per image sounds small. Across 3,000 images, it becomes 50 hours.Illustrative arithmetic, inspired by Nuuly’s account of time saved.

That calculation explains the appeal to busy studios better than a futuristic demo. The service can release time for lighting, styling, and creative work. Nuuly’s account credits its connected Creative Force studio software, too: the handoff was already part of the working system.

The price of a finished morning

Pixelz occupies the space between software and a production department. Customers include retailers, brands, and commercial studios. Its customer page claims more than 30,000 brands and retailers have trusted the service; that is a company-reported reach figure, not a current subscriber count. Public stories include ASICS, BESTSELLER, Trove, and Nuuly.

The Professional plan currently lists $95 a month, with editing starting at $0.95 per image. It advertises next-morning delivery and guaranteed capacity of 200 images a day. Enterprise agreements use custom pricing and service commitments; expedited turnaround can start at three hours.

In Pixelz’s BESTSELLER interview, sourcing manager Anders Tvede Pleth discusses total ownership cost, tooling, and training speed. That is the sensible comparison with freelancers, an internal team, or another outsourcing agency. Count the internal review time and missed handoffs as well as the edit. A low-volume, simple job may need far less infrastructure.

The model is new. The review remains.

The same production logic now extends beyond conventional retouching. Pixelz launched synthetic AI fashion models in April 2025 and introduced Digital Twins that year. The distinction matters: synthetic models are generated people; twins use the likenesses of real, consenting people. Its announced partner, talent agency Le Troch, brings model relationships and licensing into the arrangement.

Its 2026 descriptions are candid about the work behind the output. Digital Twins require multi-angle product photography, a creative brief, agreed usage rights, and review. Synthetic-model images undergo an aesthetic check before finishing. Unconvincing anatomy, inaccurate garments, or poor lighting can send an image back for generation. A plausible picture still has to represent the merchandise.

AI Image to Video adds motion from existing stills. Separately, Pixelz edits photographed footage; its April 2026 platform update describes mobile uploads and side-by-side video review. For a retailer, the attraction is handling more formats through a familiar production relationship. The condition is equally clear: usable inputs and someone accountable for approval.

There is capital behind the expansion, too. Adelis Equity Partners became the majority owner in December 2021, while management retained significant ownership and stayed in place. The stated plan was international growth and further development of the platform. The investment announcement also anticipated extending the business into additional media types.

Copy the handoffs

People remain part of the operating model. Pixelz has described training and hiring collaborations with REACH and Digital Divide Data. Its published values emphasize data, relationships, and ownership. Those principles fit a business where an instruction must travel across time zones without losing its meaning.

The transferable idea is concrete. Map where an image waits. Define which decisions repeat. Give ambiguous judgments a named owner. Measure approval rates alongside speed. Then test a manageable slice of work before moving more. Trove’s jacket supplies the final check: does the finished image tell the buyer the truth?