Interview With Ivan Altsybieiev and Dima Shvets, Co-Founders of Reface
Read the interview with co-founders of Reface. We talked about the AI behind the face swap, ethics and privacy, and, of course, memes!
Join the DZone community and get the full member experience.
Join For FreeYulia: Hi, guys! Thank you for agreeing to meet me today. Let’s start with a simple question. How would you describe Reface in a few words?
Ivan: Everything started with a face-swap. We created state-of-the-art technology and made it possible to provide it to more than 180 million people. Reface is an app that allows users to swap faces in images, GIFs, and videos in seconds and with high accuracy. Moreover, in 2021 we developed more AI tools for personalized content, which allow us to animate faces, place faces on objects, and voice over them), empowering users to generate completely new and viral content.
Yulia: What makes you different from other projects who work with face-swap technology?
Ivan: Everyone who tried to do this before us either lost in speed or in quality. When face-swap technology first appeared, we started to think about how we could wrap it up and present it to the maximum number of people. Face swap existed before us, of course, but I think it’s our big accomplishment that we envisioned how it can actually be used by people ― to make memes. For example, Facebook had it for some time but people didn’t have an opportunity to try it out. We actually gave people this opportunity: now they can change faces on photos, GIFs, and videos.
Yulia: How did Reface start?
Ivan: At first, we started with a website where you could switch faces on images, then the concept of the Reface app appeared. We decided to try out GIFs. The technology wasn’t perfect back then but it was the moment when Elon Musk tweeted about us. We thought that if Elon Musk noted our technology as worthy, then we should definitely continue working on it.
For about a year, we have been working on improving the technology. In summer 2020 we reached a point of hyper-growth. With a team of 25 people, we invented a format of content that provoked emotions in people and created a wow moment: the quality was high and people were shocked that it was even possible. But we decided not to stop there. People don’t need to change faces every day, right? There aren't many user cases here.
Yulia: So what are the new use cases that you see for Reface?
Ivan: We’re a very fast-growing team. We had just 30 people when we became popular in 2020 and now it has reached 200 employees. We believe that our ML engineering team is definitely the strongest one in Ukraine and probably one of the strongest in the whole of Europe. We did and still do our best to attract talented developers. And now we have strong expertise in ML and release benchmarks one after the other. Face animation, lip synchronization are just some of the examples. And we started thinking about memes.
Yulia: How can Reface be useful for meme creators?
Ivan: Well we believe that memes are the meta-language of the Internet. Think about TikTok, for example. Videos there are very engaging because they’re all about you. In TikTok, you film yourself, making yourself part of some narrative, such as a funny story, for instance. We believe that there are many creative people in the world who don’t want to be actors, they want to be directors. Our mission is to empower content creators via ML tools. We create these tools, wrap them up, and make them easy to use.
Dima: If these people want to create content but they don’t have tools, it becomes very hard for them and reduces their competitiveness in the creative market. They are super invested in what they do and want to create powerful content, but they can’t, of course, compete with so-called fast food, like Instagram bloggers with million followers and their teams.
Yulia: What is the future of Reface?
Ivan: We want to build a social platform that is based on memes. People can create an account and start creating memes. ML today can divide a video into a set of assets: face, voice, background. We also call it decomposing the content. For example, today Dima was giving a speech (at Web Summit 2021), and I was streaming him. But we could also make it so that Elon Musk sits next to him and looks at him. Our vision is that memes are a new way of communicating. And Reface makes it easy to create them. Remember SoundCloud? It was a place where people could not only listen to music but also create it. Our platform would be the same for memes. Creators will make content, and people will come to consume this content.
Yulia: Reface is a very complex product. Have you used any third-party instruments to make your life easier, so-to-say?
Dima: That’s the thing! It’s all in-house, all patented. ML is developing fast, and you must keep an eye on everything that is going on. And our team is, of course, monitoring new technologies that appear. But everything that Reface uses is proprietary.
Yulia: Are you planning on any new features? How do you approach new feature development?
Ivan: It’s complex. For us, it’s important to create a technology that causes strong emotions and serves the needs of creators. To achieve that, we communicate with the community and just ask them what they need. For example, they say they need a green screen. We can give it to them via ML so that our creators don't have to spend hours editing their videos in unwieldy third-party software.
Yulia: Have you ever faced unethical uses of Reface?
Dima: Great question. When we were starting we put ethics as one of the core principles of Reface. We are very invested in the topics of compliance and responsible consumption. We even have a dedicated consultant who specializes in the questions of product ethics and safety. She communicates with dedicated departments of Adobe, Facebook, and other Silicon Valley companies. Nobody actually knows today how to approach these topics the right way. But together we’re trying to resolve these issues. Another thing has to do with the product itself. On the platform, we use powerful tools for content moderation, labeling, and violation detection. But you can’t actually use Reface to mislead, and it was done on purpose. A unique tool that we are working on is invisible watermarking. This is encrypted metadata that helps to identify the user. Even if something theoretically happens, we will be able to detect the person.
Yulia: How do you approach data privacy?
Dmitry: We don’t store personal data. This is one more thing that is special about us.
Ivan: When you upload a photo or video to the platform, it calculates a certain numerical vector. This data cannot be used to identify your face. Coordinates of your face are necessary for face swap. And then in 24 hours, your photo is deleted from the server. Now we explain to people what embedding is and ask them to give consent. We’re trying to be super transparent about how we treat our users’ data.
Yulia: Thank you so much for the interview! Long live the memes!
Opinions expressed by DZone contributors are their own.
Comments