Matthieu Rouif on PhotoRoom's Focus, Mobile Growth and AI Design

PhotoRoom co-founder Matthieu Rouif tells Harry Stebbings how a frustrating image-editing task helped inspire the company. While working at GoPro on a video app, he needed promotional artwork quickly but found the editing tools too complicated for the result he wanted. He saw an opportunity to make image creation easier.

Rouif says earlier startup experience taught him to focus. At PhotoRoom, he and co-founder Elliot narrowed from photo and video to photography, then toward people selling online. He reports that each narrowing improved engagement; he recommends waiting for clear product-market fit and customer pull before broadening again.

He values rapid execution because each release can create a learning loop. But speed alone is not learning: before shipping a small version, the team should state what it hopes to learn and the hypothesis it wants to test. Otherwise, he says, it is easy to interpret any result as a success.

The founders tested early PhotoRoom versions in a McDonald's, offering to pay for people's food in exchange for trying the app. They asked about photography habits and watched users go through onboarding. Rouif says the exercise exposed assumptions so often that he was eventually barred from the nearby restaurant for disturbing customers.

One user spent minutes choosing a selfie because they thought it would become an account picture; Rouif had expected a simple demonstration. The team would change wording or a feature and return to test quickly. His lesson is to observe what people actually understand rather than explaining how the product is supposed to work.

Stebbings asks how to respond when investors dismiss background removal as a feature rather than a product. Rouif says the initial feature can reveal a deeper set of needs: focusing attention, blurring, templates and creating thumbnails. He favors making a narrow experience highly useful before expanding around it.

Stebbings cites PhotoRoom reaching $20 million in annual recurring revenue on $2 million of invested capital. Rouif says the team did not simply refuse to spend: it reinvested in growth and paid for good tools, but preferred small teams of strong individual contributors. The financial figure is the interview's historical characterization, not a current result.

He credits mobile app stores with global distribution that a small product team could access without building a large sales organization, while acknowledging their fees. He says upfront subscription payments supported quick acquisition payback; he declines to disclose exact customer-acquisition costs. These are his company's reported economics, not a promise for all mobile apps.

PhotoRoom often works with freelancers, sometimes starting with a short period of collaboration so both sides can assess the work and culture. Rouif says the company seeks skilled people who value autonomy. He also describes three cultural principles: challenge ideas to make progress, learn quickly through shipping and take pride in craft.

Growth can plateau between periods of rapid expansion, Rouif says. His response is to revisit customer needs and look for adjacent problems rather than assuming every week should rise. He says the company discussed growth frequently but wanted employees motivated by the mission as well as the graph.

PhotoRoom measures engaged people by how many users export photos regularly, with weekly active use important for its e-commerce customers. Rouif prefers metrics that count people doing something meaningful rather than only a total number of exports.

The product team can focus on an online seller's needs without excluding other users, Rouif says. PhotoRoom had users beyond its core persona and an API that others could use in different ways. The distinction is between a focused development roadmap and a restriction on who can access the product.

Rouif says early growth came from word of mouth, mentions by creators and listings carrying a PhotoRoom watermark. He recalls reaching product-market fit shortly before COVID-19 increased the need to sell online and made in-person photography harder. He recognizes the timing helped the business.

The free product's watermark creates awareness, while paid users can remove it. Some users bypass the watermark by taking screenshots; Rouif says the company measures this but prioritizes creating value and growing rather than trying to block every workaround. He expects some engaged free users eventually to pay, without claiming all will.

Asked about churn, Rouif says customers sometimes no longer need the service or move to another app, but declines to give PhotoRoom's annual churn rate. Stebbings raises a hypothetical 25% to 40% annual churn range for consumer subscriptions; it is not a disclosed PhotoRoom metric.

Rouif argues that mobile reaches a very large audience, allowing a broad group to try a product while the team focuses on deeply engaged users. His figures about potential global mobile reach are his estimates, not a count of PhotoRoom customers.

On competition with Canva and Adobe, Rouif positions PhotoRoom around entrepreneurs and small businesses who need accessible, professional-looking photos on a phone. He predicts room for multiple large businesses serving different needs, including an ambitious future revenue goal for PhotoRoom. That is his view as the company's CEO, not a proven market outcome.

PhotoRoom also offers an API and web experience. Rouif cites a partnership around the Barbie film that let people create posters using its technology, and a food-delivery example where restaurants can produce more professional images. He presents these as extensions of the same goal of helping businesses create visuals.

His concern is using new technology merely to reproduce an old product. He recalls an earlier idea for smartphone postcards and contrasts it with sharing pictures digitally. As AI changes what is possible, he wants PhotoRoom to rethink the experience rather than simply attach a new model to an unchanged interface.

Rouif says data quality matters and model size is a trade-off between quality, speed and cost. A specialized model can be smaller and faster for a particular job, he argues. He stresses speaking with customers to learn which trade-offs they actually value; performance claims in the conversation are company assertions.

He believes product and model development should inform each other. PhotoRoom can start with existing or open models to explore value, then train and adapt its own when that improves its product. In his view, commoditized building blocks can speed innovation without eliminating the value of technical expertise.

Rouif thinks incumbents and startups can both benefit from AI, but that established products face pressure to protect their existing users and workflows. A new entrant can serve people who previously could not afford professional help. His copilot-versus-autopilot analogy is a strategic argument, not a prediction that all existing products will disappear.

He believes many people do not want to type elaborate prompts for routine tasks, especially on phones. The right input depends on the job: a photo, voice or direct manipulation can convey information that text alone would make cumbersome. He expects interfaces to change, not to become unimportant.

On pricing, Rouif favors aligning what a user pays with the value received. He sees seat pricing as convenient in some settings and imagines more outcome-linked arrangements for services that automate work. The panel does not establish a single future pricing model.

Image generation can require costly GPUs, but Rouif expects research and optimization to lower cost and latency. He says PhotoRoom trains models partly to serve users faster. His anticipated timeline and speed claims reflect the state and expectations of the company in 2023.

Asked whether venture investors are overexcited about generative AI, Rouif sees a major opportunity while allowing that hype can exceed near-term reality. He does not endorse a precise estimate for how much venture capital will be lost. He distinguishes research-heavy efforts from PhotoRoom's stated emphasis on user feedback.

PhotoRoom uses observations of customer behavior to make model and interface decisions, Rouif says. In his account, mobile users do not want to wait or write long prompts, so the team works on fast image generation and a simpler interaction. The under-one-second performance he cites is a company claim at the time.

He says watching his children use apps has reinforced the importance of accessible design. They can dictate questions or move objects in an image without first learning professional editing concepts. These are personal examples, not a study of all children.

Rouif expects advances in software to let smaller teams generate more revenue and says PhotoRoom tracks revenue per teammate. He cautions against using that measure to refuse necessary hires: the company had recently doubled its team. The possibility of a ten-person company producing $1 billion in revenue is his prediction, not an achieved result.

After a gap in the automatic captions, Rouif discusses parenting and leadership. He believes leading by example and helping children grow have informed his management style. He describes balancing intense work with time for his children as difficult, rather than offering a formula.

In quick-fire questions, Rouif says prompts will remain useful for experts even if many users prefer other interfaces. He is optimistic about OpenAI and thinks AI products are still early in their development. These statements are his opinions at the time.

He says he changed his mind about creativity requiring everyone to work in person. PhotoRoom became hybrid during COVID-19; the team was distributed across Europe and met in Paris monthly. He values diversity and access to talent, while noting that some fine details of craft can be harder to resolve remotely.

Rouif describes a culture that defaults to public discussion instead of private messages, with exceptions for sensitive matters such as compensation and performance. He believes open threads allow more people to contribute and make challenges visible; it is PhotoRoom's practice, not a prescription for every organization.

He credits Y Combinator with helping him improve storytelling and think about scaling, while acknowledging that participants in the program already select for ambitious goals. His claim that it put the company on a path to $1 billion in revenue is a personal aspiration, not an achieved milestone.

Looking toward 2033, Rouif wants small businesses and entrepreneurs to create professional visuals without hiring a large agency. He imagines AI handling more of the iteration so that a restaurant owner or online seller can focus on their own work. That future vision closes the interview.