Avatars, especially with the metaverse in the wings, will be a big part of our future communications. But often these are far from their original and require hours of loving customization until they fit. Of course, there are many users who enjoy this very process, but for those who lack the time and creativity to do so, Google now offers a new option with the PHORUM AI model.
Google-trained artificial intelligence can create stunning 3D avatars from a single photo. This makes previous setups with multiple cameras, cumbersome scanners or depth cameras obsolete – because not only are they very expensive, they continue to be highly flawed. Google’s AI is expected to not only make the process easier and less expensive, but also vastly improve the quality of 3D avatars. Google’s PHORUM is an end-to-end trained AI system that not only factors in position and scale, but also takes into account shading, brightness of the body, and many parameters that were previously difficult to calculate.
The model was trained with over 200 scans of people in different positions, outfits, accessories and in front of different backgrounds, resulting in a data set of almost 200,000 images. This was then used to define grids, which also produce impressive results with other images.
Pictures and Video: Google