The Paris-based startup has raised $43 million in Series B funding and announced a fundamentally new model faster than other AI models in generating images, according to Forbes USA .
In 2018, Matthieu Rouif, a product manager at GoPro, hit a roadblock. He needed to create visuals for the camera maker’s latest video editing app and his designer was on vacation. He spent the afternoon scrolling through Photoshop in frustration. At the same time, in the next room, the company’s AI team was automating parts of photo editing as part of internal research. An idea struck: «We should give this to everyone who edits photos on their smartphone,» Rouif told Forbes.
The experience inspired Rouif to leave GoPro and start Photoroom, which offers AI-based photo editing software. The startup’s 30 million active users range from jewelry makers in Kenya to fashion resellers on Poshmark and restaurant owners on DoorDash and Wolt.
Rouif told Forbes that since the company’s founding in 2019, Photoroom users have modified more than 5 billion images with Photoroom, using the app to enhance their product photos by adding AI-generated backgrounds, improving photo resolution, and generating photos product from scratch with AI: «We realized this wasn’t just a product manager problem at GoPro, but a need for at least half a billion people who sell something every day or week and they need good photos to sell,” said Rouif, who sold his last startup to GoPro in 2016.
Photoroom announced last week that it had raised $43 million in Series B funding at a $500 million valuation. London-based venture capital firm Balderton Capital and Aglaé Ventures, an investment firm backed by LVMH CEO Bernard Arnault and his family, led the round, with participation from Y Combinator.
The new round brings Photoroom’s total funding to $64 million. With more than 150 million app downloads and a subscription-based business model, the Paris-based startup has surpassed $50 million in annual recurring revenue, according to Rouif.
Photoroom has also attracted the attention of brands such as Netflix, Lionsgate and Warner Bros, who have used the startup’s API to promote movies and shows including Barbie and Black Mirror.
In October 2023 Photoroom partnered with Universal Music Group-owned label Republic Records to create a custom selfie generator for Taylor Swift’s album 1989, which millions of fans used to create a cover of album with their own faces.
Photoroom initially gained traction in 2020, the same year it was accepted into Y Combinator. During the pandemic, entrepreneurs rushed to produce online catalogs of their products and, without access to professional photographers and photo studios, turned to photo editing tools like Photoroom. Before generative AI tools became popular, the startup’s most popular tools were a background remover, a tool called «magic retouching» that removed unwanted objects from a photo, as well as a function that could blur backgrounds in two seconds. When more advanced AI tools became available in 2023, the startup expanded its offerings to include fully AI-generated backgrounds, where users could create background visuals from scratch via text prompts — now the most used feature of Photoroom.
Along those lines, the startup has also released its own model, Photoroom Instant Diffusion, whose speeds have increased by 40 percent and is faster at generating images and backgrounds than other text-to-image AI models, the company said. (Midjourney, Photoroom’s competitor, takes 30 seconds and Dall takes 15 seconds, while Photoroom takes just one second to generate an image from scratch, Rouif said.)
The model is trained on a curated database of millions of images, some sourced from image providers and others purchased directly from artists. It’s able to create images faster through a «custom architecture» that incorporates more computing power in a given amount of time and is optimized for the latest Nvidia GPUs, co-founder and CTO Eliot Andres said: «The main distinguishing factor in this model is speed. Our users want to save time, they have busy lives, they have businesses, and we allow them to generate backgrounds and images extremely quickly.»
Rouif and Andres, who met through former colleagues, said they initially encountered some technical challenges in trying to improve the quality and accuracy of their AI model. Mostly, business owners don’t want their product details to change during editing. They just want the background scenery refreshed, Andres said. But sometimes the AI couldn’t properly differentiate between foreground and background and ended up changing small details of the object itself.
“People were disappointed because it didn’t match the products. And so we worked hard to maintain that,» Andres said. «The hardest part was keeping the contours of the object.»
Rouif said he plans to use the new funding to improve the technology, investing in more GPUs and data to train the AI model, as well as expanding his 60-person team. «Our mission is to make Nike-level, LVMH-level photography accessible to anyone in the world,» Rouif said.