Hitting 250M+ downloads with a mix of virality and cutting-edge tech

Dima Shvets is an angel investor and a cofounder of Reface, an AI-powered face-swap app that hit 250M+ downloads. He attributes the app's success to the viral nature of being the first real-time, in-app face swap technology.

Here's Dima on how he did it. 👇

Leaving funding for building

I spent five years running Imperious Group, a seed-stage VC fund, before jumping into the founder seat myself with Reface. Leaving a venture fund to build a startup was a questionable financial decision and a big leap of faith. But sometimes, those leaps pay off.

As a VC in Central and Eastern Europe, I was constantly hunting for globally scalable deep tech, and I saw two things.

1. Ukraine has world-class engineers, but most were stuck in outsourcing rather than building.

2. Deep tech has defensibility — if you had the right team, you could build something hard to replicate.

Then some of my colleagues introduced me to my Reface cofounders Roman Mogilnyi, Oles Petriv, and Yaroslav Boiko. And that’s when my part of the story at Reface began.

At that time, Generative Adversarial Networks (GANs) were emerging as a groundbreaking technology in the AI field. When the team showed me their early face-swap prototype, it clicked. 

So, I bet on the team, the tech, and the potential to define a new market.

Making the first real-time face swap app

At the core of Reface's appeal was its own face-swap model, based on GAN technology. We built an ML infrastructure that allowed us to parallel multiple AI models while keeping them lightweight and delivering consistent results.

The app used AI algorithms to accurately map facial features, enabling seamless face swaps in videos and images. This set Reface apart from competitors as one of the first teams to make real-time, on-device face swaps work.

If you ask my cofounders, we might argue about the timeline.

But from my perspective, it took about a year to go from a cool tech demo to a viral product — and a $5.5M round from a16z.

We started by launching a website where users could swap faces in images, which garnered attention and positive feedback.

Encouraged by this response, we expanded our technology to handle GIFs and videos, leading to the development of the Reface app.

At the moment I joined, we had:

•  A face-swap prototype that worked on photos (not videos).

•  The ability to hire our first full-time engineers.

• A plan to raise angel funding and turn the prototype into an actual product.

We experimented with different models for growth. We tried B2B, collaborating with Amazon Video, Universal, Warner Bros., Saatchi & Saatchi, other content holders, creative productions, and a few fashion brands. We hit $45K MRR, but B2B didn’t scale as fast as we envisioned.

So eventually, we went all-in on B2C, focusing on entertainment, content creation, and virality, while continuing to work with B2B partners as part of our marketing efforts rather than a separate business track.

We bet on a simple, fun, potentially viral use case — and it worked.

Harnessing virality

We hit 100k downloads in our first month. We started with a Product Hunt launch to bring in early adopters. We ranked 3rd in the weekly rating and won the Product Kitty Award 2020.

In August 2020, we became the #1 app in the US App Store, beating TikTok, Netflix, Hulu, and Disney+.

What worked for us:

• The novelty effect. AI-generated face swaps had never worked this seamlessly before.

• Viral mechanics were built into the product. It wasn’t just a tool, it was entertainment. More on that later.

• Celebrities and influencers adopted it organically. Instead of chasing top-tier celebrities directly, we leveraged secondary celebrities — friends of A-listers who were easier to reach. And we tagged them relentlessly. The more they reposted our tags, the bigger names started noticing Reface. As a result, Musk, Snoop Dogg, Ariana Grande, Justin Bieber, and others posted Reface organically! No sponsorships, no paid promotion — just strategic outreach.

• The press and PR flywheel kept the momentum going. Our in-house PR team secured features in Mashable, Forbes, The Verge, TechCrunch, Vice, Hypebeast, Business Insider, and many more. And once celebs started posting, the media really ran with it. Don’t neglect PR, it pays off. Find someone early on who can shape your story for the media.

• We used a freemium model where people could try it for free, get hooked, and pay for premium features.

• Content + AI = virality. The best marketing wasn’t ads, it was people sharing their face swaps.

And remember, your product’s utility defines its growth curve. Some products are naturally retentive (habit-forming). Some are hype-driven (need constant virality). Know which one you’re building.

Be careful with paid marketing.

It’s easy to grow vanity metrics with ad spend. But performance marketing can sometimes send the wrong signals and distance you from what your audience really wants.

Organic growth keeps you connected to your audience’s needs. If you only grow when you pay, chances are, you’re just burning money.

Viral mechanics

As I mentioned, viral mechanics were built right into the product. Here are some examples:

1. Watermarks: One of the most effective viral mechanics was the dynamic watermark — a less obvious move at the time. We made it impossible to crop out, ensuring every shared video carried the Reface branding. Beyond growth, this also acted as a content moderation tool, preventing misuse for NSFW content or deceptive deepfakes. Everyone knew the video was generated in Reface, reinforcing trust and authenticity.

2. Multi-face swaps: We noticed users manually creating multi-face swaps, so we built the feature directly into the app. This simple addition tapped into group dynamics. A single-face swap was fun but often shared one-on-one, whereas a multi-face swap was perfect for group chats, making it an inherently social experience. This shift boosted peer-to-peer sharing, even more than public social posts, driving massive organic growth.

3. In-App content: As an entertainment-focused product, we stayed on top of cultural trends: Super Bowl, Valentine’s Day, meme-worthy moments — we created content around them. We also monitored organic traffic spikes and capitalized on them. If a TikToker started a wedding face-swap trend, we made sure users could find more wedding-themed content inside the app.

Virality pains

The hardest part was scaling AI when Elon Musk retweeted a Reface video. We were suddenly handling millions of downloads overnight.

Our backend wasn’t ready. We had two weeks of chaos where our engineers had to figure out how to scale to over one million DAU without breaking the product.

We scaled our backend, optimized AI processing, and didn’t lose a single viral wave.

Parting advice

I have three pieces of advice.

1. Your support system is everything. Focus on your health, family, and mental clarity. Startups are marathons with constant sh*tstorms — your endurance defines whether you survive.

2. Every decision is a numbers game. It’s impossible to build something without making mistakes, but if your good decisions outnumber the bad ones, you’re winning.

3. Hiring the wrong people costs you more than not hiring at all. Every startup struggles to scale the team as fast as the product. In hindsight, the fewer people on the team, the better. Try to automate and optimize your workflow using AI to keep the team small. Larger teams lead to blurred values, a diluted team vibe, and a higher risk of hiring toxic or inefficient people.

And here are some books that helped me along the way.

• The Hard Thing About Hard Things by Ben Horowitz is a must-read for anyone in startups. It’s brutally honest about what it takes to build a company.

• Zero to One by Peter Thiel gives a great perspective on building monopolies and thinking outside of competition.

• The Daily Stoic by Ryan Holiday and Stephen Hanselman helped me develop an entrepreneurial mindset and stay resilient through the ups and downs.

What's next?

In 2023, I stepped away from my CEO role at Reface to build a new AI infra dev tool. But it’s still in stealth. Stay tuned.

 Follow me on, Twitter, Linkedin, Threads, and Substack.