The episode features Romain Torres, co-founder of ARCAds.ai, a platform that helps businesses generate video advertisements using AI avatars without expensive agency fees. Romain is joined by host Nathan Latka to discuss the company's rapid growth trajectory.
ARCAds.ai launched in January 2024 and has scaled to $10 million in annual recurring revenue with over 6,000 paying customers, all while maintaining a lean team of just eight people and remaining bootstrapped and profitable.
The conversation covers their viral growth moment in March 2024, the technical infrastructure behind their AI video generation, and their three primary growth channels: enterprise sales, paid advertising, and influencer marketing.
Romain shares insights on building product-led growth, managing hundreds of ad variations with a small team, and the strategic decision to focus on enterprise customers paying six-figure annual contracts.
From Zero to $10M ARR in 20 Months
ARCAds.ai reached $10 million in annual recurring revenue as of November 14, 2025, growing from first dollar revenue in January 2024. The company now has over 6,000 paying customers, double the count from earlier in the year.
Initial validation phase involved manual service delivery without a dashboard. "We were reaching out to existing brands and told them, okay, we are going to create ads for you, UGC video ads... using AI" - Romain. Revenue during this phase was approximately $5,000.
The company maintains "very healthy profitability" and "super healthy margins" while growing over 20% per month, with Romain describing margins as "a lot" when pressed for specifics.
Viral Growth and Product Breakdown
March 2024 viral moment drove revenue from €5,000 to €64,000 overnight when a customer's AI-generated video went viral on Twitter. "I was scrolling in the Slack channel with all the Stripe notifications. We got like hundreds of sales in one single night" - Romain.
The viral spike immediately exposed critical product failures. The platform was "completely broken" and required a full month to rebuild major components, forcing refunds for all new customers before relaunching in April 2024.
Growth temporarily flattened when Romain shifted focus from marketing to building the sales team. "When you don't do marketing and you don't promote your product, it doesn't really grow as fast" - Romain. Organic growth continued at 10% monthly during this period.
Technical Infrastructure and AI Model Strategy
ARCAds.ai operates as a user experience layer on top of foundational AI models from OpenAI, Google, Meta, and Chinese companies like Kling and ByteDance. The platform has over 1,000 AI avatars in its library.
The platform automatically selects the best AI model based on user prompts. For avatar generation, it uses either Google's Imagen or SeeDream from ByteDance (TikTok's parent company).
Custom model development focuses on specific use cases that large foundational models don't address. "We'll build sometimes custom models that are not provided by these providers... that focus on very specific use cases that they won't tackle" - Romain.
Users can create custom avatars by describing appearance or uploading product images. The platform demonstrated generating a "young male who just completed a Hyrox" holding a Nike shoe, with the AI correctly understanding the Hyrox environment context.
Enterprise Sales and Usage-Based Growth
Highest annual contract value (ACV) reached six figures, with Romain stating "it's only going to grow." Enterprise clients include companies like Adidas, Nike, and Samsung.
Usage-based pricing drives enterprise growth as companies shift from 0% to 100% AI-generated ads. "Some companies are 100% AI generated ads... the usage is growing more and more and more" - Romain.
Sales team expansion from zero to two people, combined with conference attendance and direct outreach, became the primary driver of recent growth from €500,000 to €800,000 monthly revenue.
Product-led growth remains foundational, with Romain emphasizing that enterprise customers need sales touch to convert despite the self-service product. "They have much higher LTV, they are much better clients, but they need to be sold the product" - Romain.
Managing 210+ Ad Variations with 8 People
The company currently runs 210 active ads on Meta as of November 2025, managing creative iteration through their own platform to create a direct feedback loop.
Ad testing system starts with idea collection in Notion, then rapid generation of variations. "Before AI... you could test like, I don't know, five ideas every single month. But then AI comes and you can go and say, I'm going to generate video variations" - Romain.
Successful ads are translated into multiple languages and tested with different avatars. The team A/B tests faces, scripts, and languages, with frequently appearing faces indicating strong performance.
Profitability focus guides ad spend decisions, though Romain declined to share specific budget allocations between paid ads, influencer marketing, and API costs due to competitive concerns.
Three Primary Growth Channels
Top three growth channels driving recent expansion: enterprise sales team, paid advertising using their own platform, and influencer marketing partnerships.
Paid ads serve dual purpose of revenue generation and product development. "We'll not only grow the business doing that, but we'll also grow our knowledge about how the product should look like and the features we should build" - Romain.
Influencer marketing expanded from organic content (Romain's Twitter posts and customer viral videos) to paid partnerships, though specific spend and influencer counts remain confidential.
SEO represents untapped opportunity with only 5,300 organic clicks per month to arcads.ai according to Ahrefs. "We did nothing basically in terms of SEO, but just like organic content" - Romain. The company is now hiring for SEO expertise.
Paid Ads Strategy and Common Mistakes
Biggest mistake founders make is insufficient testing volume and duration. "People will put like a hundred bucks and say, it's not working. They will like just test one ad with 100 bucks and they will basically waste $100" - Romain.
Learning paid ads requires sustained practice, compared to learning to swim at 40 years old. "You have to go to the pool every single week until you get better... That's the same thing for paid ads" - Romain.
AI enables dramatically higher testing volume, making it easier to find winning ad combinations through iteration rather than single attempts.
Hook generator tool at arcads.ai/hook-generator helps users create effective opening lines by describing their product, addressing the challenge that most founders struggle with script writing.
Building Moats Through Product Excellence
Product-led growth remains the number one acquisition channel despite Romain's marketing focus. "If I had to only pick one, I will focus on products because... if you have the best... people will find you" - Romain.
Co-founder division of labor: Dylan handles all product development while Romain manages marketing and sales exclusively, enabling deep focus in each domain.
50-50 equity split between co-founders who have worked together for 10 years since age 18. "The thing I'm most happy with when it comes to my career is finding Dylan and working with him early on... that makes the journey 20 times better" - Romain.
Vision for AI automation: future platform will automatically describe products, generate scripts, select actors, test in ad accounts, and iterate based on performance data without human intervention.
Market Position and Future Vision
Romain predicts "every single ad will be AI generated in like three years," positioning ARCAds.ai at the intersection of technology and creativity similar to Apple.
Would decline a hypothetical $150 million all-cash acquisition offer. "We are probably building one of the most exciting thing one can build... there is nothing more exciting we could be building" - Romain.
Next growth priorities focus on enterprise sales expansion and SEO, with the company actively hiring for SEO expertise to capture organic traffic opportunity.
At 30 years old, Romain has spent his entire business career working with co-founder Dylan, building multiple ventures together over the past decade.
From SaaS Interviews with CEOs, Startups, Founders. Get a note like this from every new episode.