.webp)
Photoroom has launched Visual Agents for enterprise product catalogues, a system designed to manage AI image and video production for retailers and marketplaces.
The product builds workflows around incoming catalogue images, selects editing models for each task, checks the output against the original product, and retries flawed results before sending borderline cases to a human reviewer.
Its release comes with benchmark data that highlights a broader problem in AI-generated commerce imagery. In tests of 4,250 AI-generated images covering 850 real products across clothing, footwear, bags, jewellery, and accessories, the best image-editing models passed a full product fidelity check only 29% of the time.
Logos were distorted or unreadable in one in five outputs in those tests. Photoroom said its Fidelity Layer, used throughout the Visual Agents process, raised the pass rate to 38.2%.
Quality checks
Photoroom presents the issue as less about access to AI models and more about connecting tools, vendors, and review processes that can handle large catalogues without introducing inconsistent results. It says many businesses still rely on manual quality control because one model may succeed on one image and fail on another, even for the same product.
Company engineers found that repeated attempts could improve results without changing the underlying model. That led to the retry loop inside Visual Agents, which is meant to catch recoverable failures before human reviewers get involved.
Photoroom said several large marketplaces have already built internal systems to score and filter AI-generated product images. It argues that no equivalent commercial option has existed for companies without the resources to build those systems themselves.
Early deployments cited by Photoroom span marketplaces and retail platforms with large volumes of product photography. One unnamed consumer marketplace ran more than 1.5 million automated quality checks in three weeks and corrected more than 90,000 images after defects were found.
A food delivery marketplace processed 8,200 catalogue images in one overnight run, with 99.9% completed automatically and seven sent to a human reviewer, according to the company. A fashion resale marketplace recorded a 2.2% lift in gross merchandise value over seven days against a control group, with results ranging from 2% to 5% by product category, using Photoroom's Create & Transform API, which now runs inside Visual Agents.
Review burden
Photoroom is focusing on reducing the labour involved in visual review at scale, particularly for retailers and marketplaces that process large numbers of product listings each day.
"I have run imaging operations that processed millions of products a year. The part nobody sees is the review queue. Every image that needs a person is a minute of somebody's day, and at catalog scale those minutes are headcount. Visual Agents cut that queue down to the images that genuinely need someone," said Jeff Strauss, Head of Imaging at Photoroom.
Matt Rouif described the company's view of the problem facing enterprise teams using AI for catalogue work.
"Almost every enterprise team we talk to describes the same problem. They have good models available to them, and they still cannot get a catalog they trust out the other end. That is not a model problem. It is everything around the model: knowing which one to use for which product, checking whether the result is good enough to sell, and fixing it when it is not. That work has been landing on our customers. We built Visual Agents so it lands on us instead," said Rouif.
Enterprise push
Visual Agents is available to enterprise customers, with pricing based on volume. Photoroom also offers an added guarantee for customers with stricter fidelity requirements, covering cases that still fail after the system's retry process.
Founded in 2019, Photoroom says it processes more than seven billion images a year across mobile, web, and API services. Its customers include marketplaces and commerce platforms such as DoorDash and Depop, and it operates from Paris and New York.
The benchmark figures underline the challenge for retailers trying to automate catalogue production without changing how products appear to buyers: in Photoroom's tests, even the strongest AI image-editing models passed a full product fidelity check in fewer than three out of 10 cases.