Olivier Lemarié
VP of Growth at Photoroom
Where to hear and read about Olivier Lemarié1 episode or test
Olivier Lemarié is VP of Growth at Photoroom, where he joined in 2020 after more than ten years in app marketing. The most useful thing to learn from him is how a product-first wow moment makes ads, onboarding, pricing and expansion all easier at once.
When you can put the product first in your marketing, everything is easier.
How One AI Feature Supercharged Photoroom’s Growth
Olivier Lemarié: the story and the lessons
1Olivier Lemarié's path to Photoroom
Lemarié brought over ten years of app marketing to Photoroom. He worked at Wooga in Berlin and at Molotov in France before joining Photoroom in 2020.
The team then was two founders and one iOS developer.
Product-market fit already existed when he arrived. The founders tested every week and spoke to users every day to remove friction. The early numbers were strong enough that he called it a "rocket ship".
Photoroom grew from background removal into an AI creative agency for sellers. It began as an app that only removed backgrounds. It now helps sellers list products and also market them, with social posts and logos.
Lemarié describes the shift as "AI creative agency solution for any sellers".
2How Photoroom shipped AI backgrounds first on mobile
The founders waited for the AI breakthrough and moved when it came. The first Stability model with a commercial license arrived in July 2022, and the founders saw it as the moment to build next to background removal. The CTO had worked in machine learning, and the CEO had spent almost ten years in photo and video apps.
They treated AI in photo editing as a question of when, not if, so the team was ready.
Making it work on a phone took seven months. Good generated backgrounds need good removal first, and the result had to appear in seconds. The feature launched in January or February 2023.
"on mobile you need to have the answer you can wait a few seconds but that's it"
The feature added the missing step for sellers. Resellers on Poshmark and Vinted already used batch mode for white-background cutouts. Shopify and Etsy stores wanted branded or seasonal scenes, such as a coffee cup on a wood desk.
Many users had been pasting cutouts into other creative tools. Download to finished, exported image now takes about a minute.
Four signals together told him to run ads. Product vision, tech vision, being the only one on mobile, and a happy community. Users in the Facebook group called the feature big and amazing.
The team never doubted the direction, only how far it would scale. The goal metric was subscriber growth.
3How AI backgrounds cut Photoroom's cost per install
Before and after ads show the product in one step. Before and after images have always been Photoroom's best ad format, along with a slightly controversial sentence. With the new wow moment, Lemarié says CPI dropped "by 10".
"I import a photo, I generate a background. I got what the product is about."
He treats paid growth as creative plus the signal sent to ad networks. The new feature improved both the ads and the conversion of the product experience behind them.
A product that does not convert makes marketing and growth much harder.
A weekly test loop turned strong material into scale. Test creatives Monday, read first results Tuesday or Wednesday, iterate, ship winners into the main Meta and TikTok campaigns and check ROAS.
Meta scaled fastest because small business owners are on Instagram, TikTok and YouTube alike. The hard part is catching their attention, not finding them.
A two-person team ships 400 ads a month with its own tools. The creative strategist comes from filmmaking and codes her own tools. One translates a winning ad into 20 languages, using ElevenLabs and other APIs for voice-over and adding music automatically.
She built it about a year and a half before the interview, when such tools were still new. A separate lead runs the user acquisition team.
Photoroom uses its own API to generate ads. An API for generating visuals at scale shipped about three weeks before the interview. From one product photo it can produce thousands of static ads in many languages and backgrounds.
An in-house tool built on it targets static ads for Google campaigns. Screen recordings of editing also make the before and after ads more compelling.
4Lemarié's playbook for international growth
Cheaper installs opened paid Android in low-GDP markets. Photoroom ran Android campaigns in Brazil, Thailand, Mexico and Indonesia, where subscription apps usually struggle to monetize.
The feature also reached small businesses and sellers beyond marketplace resellers, who were more willing to pay.
He sorts markets into three groups. English-speaking markets, strategic markets and regional leaders. The US and UK come first, as for any subscription app.
Strategic markets are ones where Photoroom can be first or must be big to succeed, such as Brazil, Japan and Korea.
Translated US ads work for about 80% of users. Photoroom translates and adds voice-over to winning US ads. A good US ad is rarely a bad ad in Brazil or France.
Core markets get extra local work, such as Brazil templates for Mercado Libre, its biggest marketplace there.
Japan and Korea get local agencies, not translation. Culture, willingness to pay and attitudes to SaaS and mobile subscriptions differ too much. Ideally the people are Japanese or Korean, even if they live in France.
Agencies also cover Brazil, Thailand and Mexico. The Europe-based team includes people from Brazil, China and Poland, who can quickly tell if an ad is totally off.
5Lemarié on free tiers, watermarks and pricing
Show the core value in onboarding before any login or paywall. The first onboarding was a splash screen, an imported photo and a removed background. Apple used it as an example of a great first experience.
A paywall, questions and login came later to learn about users. The rule stayed: "If you don't do that, you can't scale."
He removed the free watermark once the brand could afford it. Watermarked exports spread through Facebook communities and Reddit, and some owners even printed flyers with the logo. It was removed about a year to a year and a half before the interview.
After AI backgrounds scaled, the cost showed up in free user activation and retention, so they tested that against lost revenue. Photoroom is not Duolingo or Canva, but had enough awareness to afford the trade.
Most AI features sat behind the paywall for two years. AI backgrounds were Pro from launch, with shadows and retouching also paid, since they are the most valuable and cost real model usage. Free users get about 250 cutouts with no watermark.
By mid 2025 Photoroom was giving more away free and upselling specific features.
He bets casual free users become paying sellers later. A free user who enjoyed background removal may start a business and pay then. Sellers who earn more with the tool will pay more for it.
The risk is subscribers seeing 80% of the value free and asking why they pay. Once revenue and retention hold, product vision decides how much to give.
A weekly plan beat his fear that it looked like scamming. A teammate pushed the founder to test it. Daily users took yearly, while occasional users took monthly only because it was cheaper than yearly.
Once weekly existed those users switched, and it earned much more than monthly. It also made Pro more accessible in lower-GDP markets.
6Community and brand at Photoroom
He paid small TikTok creators a few hundred dollars. In 2021 Lemarié saw random users posting about removing a background in two seconds and asked for two more videos for a few hundred dollars. By the end of 2022, 60 to 70 creators promoted the app regularly.
They helped spread the AI backgrounds launch. His aim was reach: "it's always better to have 10,000 people who love your product", but most people should know about it.
He puts the product first instead of forcing virality. Editing for sellers is less viral than apps like avatar makers, so Photoroom catches AI photo trends and spreads them through paid ads, which lifts word of mouth.
The watermark was the early brand loop. Lemarié says brand-first growth does not work: pull in the market first, then build the brand.
7How Lemarié's team works and where he sees editing going
Photoroom runs with no DMs and everything in the open. Slack and Notion are open, including finances and what is going badly. Questions in public channels get answered in about 10 minutes instead of 4 hours.
Open channels also make a knowledge base that AI tools can query.
He sees speed as the asset that compounds most. A team that moves slower than a big competitor will die. Transparency and AI tools are his two ways to make speed compound.
He expects generated images to become the default for selling. He expects a point where generated and real images cannot be told apart, and editing may shift from tools to spoken prompts.
He admits he has no clue how this will play out.
How One AI Feature Supercharged Photoroom’s Growth
Photoroom's VP of Growth explains how adding AI background replacement to a background-removal app created a new wow moment that improved organic reach, cut paid CPI sharply and opened Android markets like Brazil, Thailand and Mexico. It also covers watermark, free tier and weekly plan tradeoffs, and how a two-person team ships 400 ads a month.
Lemarié brought over ten years of app marketing to Photoroom. He worked at Wooga in Berlin and at Molotov in France before joining Photoroom in 2020.
The team then was two founders and one iOS developer.
Product-market fit already existed when he arrived. The founders tested every week and spoke to users every day to remove friction. The early numbers were strong enough that he called it a "rocket ship".
Photoroom grew from background removal into an AI creative agency for sellers. It began as an app that only removed backgrounds. It now helps sellers list products and also market them, with social posts and logos.
Lemarié describes the shift as "AI creative agency solution for any sellers".
The founders waited for the AI breakthrough and moved when it came. The first Stability model with a commercial license arrived in July 2022, and the founders saw it as the moment to build next to background removal. The CTO had worked in machine learning, and the CEO had spent almost ten years in photo and video apps.
They treated AI in photo editing as a question of when, not if, so the team was ready.
Making it work on a phone took seven months. Good generated backgrounds need good removal first, and the result had to appear in seconds. The feature launched in January or February 2023.
"on mobile you need to have the answer you can wait a few seconds but that's it"
The feature added the missing step for sellers. Resellers on Poshmark and Vinted already used batch mode for white-background cutouts. Shopify and Etsy stores wanted branded or seasonal scenes, such as a coffee cup on a wood desk.
Many users had been pasting cutouts into other creative tools. Download to finished, exported image now takes about a minute.
Four signals together told him to run ads. Product vision, tech vision, being the only one on mobile, and a happy community. Users in the Facebook group called the feature big and amazing.
The team never doubted the direction, only how far it would scale. The goal metric was subscriber growth.
Before and after ads show the product in one step. Before and after images have always been Photoroom's best ad format, along with a slightly controversial sentence. With the new wow moment, Lemarié says CPI dropped "by 10".
"I import a photo, I generate a background. I got what the product is about."
He treats paid growth as creative plus the signal sent to ad networks. The new feature improved both the ads and the conversion of the product experience behind them.
A product that does not convert makes marketing and growth much harder.
A weekly test loop turned strong material into scale. Test creatives Monday, read first results Tuesday or Wednesday, iterate, ship winners into the main Meta and TikTok campaigns and check ROAS.
Meta scaled fastest because small business owners are on Instagram, TikTok and YouTube alike. The hard part is catching their attention, not finding them.
A two-person team ships 400 ads a month with its own tools. The creative strategist comes from filmmaking and codes her own tools. One translates a winning ad into 20 languages, using ElevenLabs and other APIs for voice-over and adding music automatically.
She built it about a year and a half before the interview, when such tools were still new. A separate lead runs the user acquisition team.
Photoroom uses its own API to generate ads. An API for generating visuals at scale shipped about three weeks before the interview. From one product photo it can produce thousands of static ads in many languages and backgrounds.
An in-house tool built on it targets static ads for Google campaigns. Screen recordings of editing also make the before and after ads more compelling.
Cheaper installs opened paid Android in low-GDP markets. Photoroom ran Android campaigns in Brazil, Thailand, Mexico and Indonesia, where subscription apps usually struggle to monetize.
The feature also reached small businesses and sellers beyond marketplace resellers, who were more willing to pay.
He sorts markets into three groups. English-speaking markets, strategic markets and regional leaders. The US and UK come first, as for any subscription app.
Strategic markets are ones where Photoroom can be first or must be big to succeed, such as Brazil, Japan and Korea.
Translated US ads work for about 80% of users. Photoroom translates and adds voice-over to winning US ads. A good US ad is rarely a bad ad in Brazil or France.
Core markets get extra local work, such as Brazil templates for Mercado Libre, its biggest marketplace there.
Japan and Korea get local agencies, not translation. Culture, willingness to pay and attitudes to SaaS and mobile subscriptions differ too much. Ideally the people are Japanese or Korean, even if they live in France.
Agencies also cover Brazil, Thailand and Mexico. The Europe-based team includes people from Brazil, China and Poland, who can quickly tell if an ad is totally off.
Show the core value in onboarding before any login or paywall. The first onboarding was a splash screen, an imported photo and a removed background. Apple used it as an example of a great first experience.
A paywall, questions and login came later to learn about users. The rule stayed: "If you don't do that, you can't scale."
He removed the free watermark once the brand could afford it. Watermarked exports spread through Facebook communities and Reddit, and some owners even printed flyers with the logo. It was removed about a year to a year and a half before the interview.
After AI backgrounds scaled, the cost showed up in free user activation and retention, so they tested that against lost revenue. Photoroom is not Duolingo or Canva, but had enough awareness to afford the trade.
Most AI features sat behind the paywall for two years. AI backgrounds were Pro from launch, with shadows and retouching also paid, since they are the most valuable and cost real model usage. Free users get about 250 cutouts with no watermark.
By mid 2025 Photoroom was giving more away free and upselling specific features.
He bets casual free users become paying sellers later. A free user who enjoyed background removal may start a business and pay then. Sellers who earn more with the tool will pay more for it.
The risk is subscribers seeing 80% of the value free and asking why they pay. Once revenue and retention hold, product vision decides how much to give.
A weekly plan beat his fear that it looked like scamming. A teammate pushed the founder to test it. Daily users took yearly, while occasional users took monthly only because it was cheaper than yearly.
Once weekly existed those users switched, and it earned much more than monthly. It also made Pro more accessible in lower-GDP markets.
He paid small TikTok creators a few hundred dollars. In 2021 Lemarié saw random users posting about removing a background in two seconds and asked for two more videos for a few hundred dollars. By the end of 2022, 60 to 70 creators promoted the app regularly.
They helped spread the AI backgrounds launch. His aim was reach: "it's always better to have 10,000 people who love your product", but most people should know about it.
He puts the product first instead of forcing virality. Editing for sellers is less viral than apps like avatar makers, so Photoroom catches AI photo trends and spreads them through paid ads, which lifts word of mouth.
The watermark was the early brand loop. Lemarié says brand-first growth does not work: pull in the market first, then build the brand.
Photoroom runs with no DMs and everything in the open. Slack and Notion are open, including finances and what is going badly. Questions in public channels get answered in about 10 minutes instead of 4 hours.
Open channels also make a knowledge base that AI tools can query.
He sees speed as the asset that compounds most. A team that moves slower than a big competitor will die. Transparency and AI tools are his two ways to make speed compound.
He expects generated images to become the default for selling. He expects a point where generated and real images cannot be told apart, and editing may shift from tools to spoken prompts.
He admits he has no clue how this will play out.