Reface became the world's most popular face swap app thanks to people’s strong desire to become anyone they want and have fun with AI. And here we are, engaging over 250M users! This year, started working with 2D synthetic content and researching hyperrealistic 3D neural modeling, the next breakthrough ML technology that would come after face swap to integrate easily into the new paradigm of the internet, metaverses, and the Web3-based platforms.
We announced Metahead, our new neural rendering technology for the instant creation of 3D objects, at one of the world’s leading events for startups and founders, Slush 2022.
What is Metahead?
Metahead is an e2e rendering pipeline that employs neural rendering to create 3D objects from 2D photos, using only pixel shaders.
Metahead allows you to render objects as complex as human heads, having only a few selfies, and get a hyperrealistic result. It is possible literally with your phone. No extra devices are needed.
What makes Metahead special?
Creating a detailed 3D representation requires enormous budgets, the efforts of a team of specialists, plenty of photos and equipment for 3D scanning, hours and months to render scanned data, and a powerful PC to finalize the result. We radically changed the previously used pipeline, commoditizing neural rendering and making the process fast and simple:
- minimum data input – just a few 2D photos
- real-time rendering
- hyperrealistic results: radically detailed textures (8K)
- full interoperability with 3D engines, editors, and platforms
- rapid integration of the result into digital worlds
- cost-effective computation
Metahead is tailored to provide a hyperrealistic customizable head with facial texture, human emotions transfer, and rendering in real-time on a smartphone or web browser with no additional camera scanning. The technology requires minimal data inputs to produce a highly detailed result and lower computation costs.
How it works
It takes a couple of pictures to create a high-resolution 3D representation with Metahead. As a result, you get a highly detailed 3D model of the face and head generated by AI, which is fully dynamic and parametric.
After learning on a huge dataset, our neural network generalizes and refines unique facial features and skin textures — such as wrinkles, birthmarks, enlarged pores, cracks on the lips, blood vessels, or even hairs. This is one of the keys to making your 3D model hyperrealistic.
The next step is to control lighting and shapes and trace the emotions. Metahead will allow you to customize your 3D model by changing lighting or head shapes, transfer real emotions and see how skin textures react to changing parameters.
Metahead is in its final stage of development and will be released soon. Share your contact info to be one of the first to access the technology and create a breathtaking hyperrealistic 3D representation in real time 🔥