Alpha3D: Shaping the Future of 3D Models for Augmented Reality

Alpha3D: Shaping the Future of 3D Models for Augmented Reality

From selling products online to constructing virtual worlds for gaming or even the metaverse – the digital world has an increasing demand for 3D models of real-world objects. But no one will create all these 3D models manually as it would take ages. 

Founded in 2018 by Madis AlesmaaRait-Eino Laarmann, and Shahab Anbarjafari, three serial entrepreneurs from the software and gaming industry, the startup Alpha3D (formerly Alpha AR) has developed its own AI engine that automates the entire process of creating 3D models of real-world objects. After winning the How to Web’s 2021 Spotlight competition, a convertible note of €350K issued by TechAngels, SeedBlink, Growceanu, GapMinder Venture Partners, RocaXSimple Capital, and Transylvania Angels Network, it raised last year an $800K seed round by Curiosity VC and ZAKA Ventures

Learn more about the future of computer-generated 3D models from our interview with the CEO, Madis Alesmaa

Why Did You Start Alpha3D?

Everything started when I met my co-founder Rait-Eino during my time in the Estonian business school, and we just knew we would be building stuff together for years to come. We co-founded several startups, in the mobile games and CRM software spaces before ultimately starting Alpha3D in 2018.

We had that light bulb moment when we learned that 3D models for AR games were created by agencies and consulting companies in a laborious manner using laser scanners, photogrammetry, and manual. We saw a huge market opportunity ahead, especially as everyone was starting to get hyped about the metaverse – and no one had solved the bottleneck of creating 3D models for it in a scalable manner. While agencies were showcasing AR apps, we set out to find a scalable way to create them in the first place. 

How Does Creating 3D Models With Machine Learning Work?

We create an AI that produces 3D models from product images by predicting the image’s depths and thereby reconstructing a 3D model of the object contained in it. This process is fully automated, i.e., we’re not constrained in terms of volume: While agencies can create maybe a couple of hundreds of items per month, we could easily do millions of them within a day – it’s just a matter of computing power.

As with all AI startups, accuracy is a major challenge: If you’re feeding your model junk food for training, it will produce artifacts. Generative adversarial networks and transformer models like DALL-E 2 can easily create images and 3D models, but their quality is insufficient to serve as training data. That’s why we’re still using images of real-world objects. 

Another challenge is that you never know which image a customer may upload. Images of home interiors are easy because that’s what people usually upload. For example, images of helicopters are a lot less common and thus harder. By working with many different customers, which involved world-leading companies, we got exclusive access to unique databases, which helped us to create 3D models at unprecedented quality and speed.  

These 3D models will then serve as content for AR, virtual worlds, gaming, or NFTs – with e-commerce having some of the hardest requirements, especially compared to gaming or metaverse applications. 

How Did You Evaluate Your Startup Idea?

We could demonstrate from a technical point of view that our AI could create 3D models faster and more accurately than agencies. Especially for e-commerce, accuracy is also critical: your 3D model should really look like the product you’re selling. By starting pilot projects with brands like Renault, Meta, or Farfetch, we could also validate our go-to-market.  

We’re still in the early days of gaming and the metaverse, and we firmly believe that content will become more shareable. Thus far, we have targeted mainly the B2B segment, but we may also explore B2C in the future: just imagine you want to sell your bike online, and you could create a 3D model on the fly using just your phone – our technology provides the backbone for that. 

What Advice Would You Give Fellow Deep Tech Founders?

Be bold because bringing real value to customers with your startup is difficult. And a startup is not about lifestyle but building a business that generates revenue – don’t forget to build a strong business case.