Sloyd 2.0

Offentlig prosjektsammendrag

The transition into automation has been going on for nearly two decades. It started with the most mundane and repeatable tasks, and it increasingly tackled more complicated processes. Now, machine are able to perform humans' most elaborate function - creating art. This revolution has only just started, and there’s yet to be an established leader in the field of 3D, neither has a dominant design emerged. Sloyd is one of a few companies with the ambition and early proven international recognition to conquer this new peak in technology evolution. Sloyd’s mission is to automate and simplify the creation of 3D models across industries. Digital 3D-modeling is a big cost driver of video games, movies, TV, graphic design, and architecture. With Sloyd the cost of producing creative 3D can be greatly reduced. The user edits 3D models just by entering text and changing simple sliders. Behind the scenes, the Sloyd engine generates the models on the fly - creating unique and customizable models. Then the user can edit all these objects at once. While generative AI has shown promise in creative fields, applying it to 3D models presents challenges. 3D models are intricate, consisting of triangles/polygons, and must be optimized for real-time settings, like mobile games. Additionally, annotated 3D data is scarce and ethically complex. Initial GenAI for 3D displays potential, but falls short in optimization and gaming suitability. Generated models lack practicality - high polygon counts, long generation times, and fused parts unsuitable for texturing/animations. Even if hypothetically this is solved, the processing power required for these operations are huge, leaving massive CO2 footprint, with high costs. We propose that combining procedural generation with ML offers the optimal path to produce usable 3D game models. Our endeavor is unique, focusing on scalability and performance, with the aim to reshape 3D asset creation.