Photoroom raises $43m Series B, bets on its own product-photo AI model

Photoroom, the Paris-based AI photo-editing startup, has raised $43m in a Series B round led by Balderton Capital, with new backer Aglaé Ventures and existing investor Y Combinator also taking part. The round pushes the five-year-old company's total funding to roughly $64m and values it at $500m, a sharp jump for a team that started out building a simple background-removal app.

The timing matters. Rather than treat the raise as a standalone funding story, Photoroom paired it with the launch of its first foundation model built specifically for product photography, signalling that the money is meant to fund model training, not just growth marketing.

Why investors are backing a model built from scratch

Most consumer apps that added "AI" over the past two years did so by wrapping other companies' models. Photoroom has taken a different route: it trains its own generative models rather than building on top of external ones. That choice is more expensive and slower up front, but it gives the company control over how its images look and how fast they can be produced, which matters for a product that promises studio-quality photos in seconds.

The new capital is earmarked for that approach. The company said it needs more GPU capacity and wants to license imagery from photographers and image providers to expand its training data. Reports around the raise also pointed to plans to significantly grow headcount through the year, part of the push to scale a team that had, until then, run notably lean relative to its user numbers.

Founded in early 2019 by Matthieu Rouif and Eliot Andres, Photoroom had already raised a $19m Series A led by Balderton in late 2022 after starting with a modest Y Combinator seed cheque. The jump from that round to a $500m valuation in just over a year shows how quickly investor appetite for applied generative AI products grew through 2023, even for a company outside the US.

The new model: Photoroom Instant Diffusion

The product news behind the funding is Photoroom Instant Diffusion, or Photoroom ID, a foundation model trained specifically to handle product photography. The goal is consistency: images pulled from different sources, lighting conditions or cameras should come out looking as if they were shot in the same studio, which is a persistent headache for online sellers juggling photos from multiple suppliers.

Matthieu Rouif, Photoroom's chief executive and co-founder, said the model is meant to remove a skills barrier rather than add one.

"The foundation model is the next step in empowering businesses to create amazing product photos without the need to be an expert at prompt engineering or photography," Rouif said.

He added that the update is also about raw output quality, pointing to sharper resolution, finer image detail and a bigger training dataset than earlier versions relied on.

From selfie app to retail infrastructure

Photoroom's pitch to investors rests on scale it has already proven with consumers: more than 150 million downloads worldwide and, at one point in 2023, a ranking among the most-used generative AI consumer apps. That reach is now the company's argument for selling into e-commerce and marketing teams through an editing API, with clients including Wolt and Poshmark.

The clearest demonstration of that enterprise pivot came from an unlikely partner: Warner Bros. The studio used Photoroom's background-removal API to power its Barbie Selfie Generator, a social campaign that let fans drop themselves into Barbie-style posters using images shared more than 13 million times.

Cameron Curtis, who leads worldwide digital marketing for Warner Bros., credited the tool's simplicity for the campaign's reach.

"Photoroom shattered all expectations. Its unparalleled ease of use became the catalyst for propelling the Barbie Selfie Generator into the digital spotlight, and captivating audiences worldwide," Curtis said.

For European founders watching the generative AI funding market, Photoroom's round is a reminder that there is still appetite for companies training proprietary models rather than reselling access to someone else's, provided they can show commercial traction to match the compute bill. The next test will be whether Photoroom ID converts that attention into paying enterprise customers fast enough to justify the GPU spend it was built to fund.