Evercoast, a provider of end-to-end 3D spatial video software, has announced the acquisition of Depthkit, a pioneer in volumetric capture.
This strategic acquisition marks a significant leap forward in the race to capture and generate 3D digital humans and spatial video, a field of immersive video content that allows viewers to see subjects and dynamic scenes from all angles on any digital device.
Today’s news also marks a key milestone for the evolution of generative AI from 2D to 3D creation, as the extensive datasets collected by both companies can fill a much-needed gap in required training data, the companies said. Evercoast has 15 people while Depthkit has three. The purchase price was not disclosed.
“Practical uses of AI have grown very fast in a very short time – and that’s largely thanks to the treasure trove of 2D and text training data available,” said Ben Nunez, CEO of Evercoast, in a statement. “But certain fields – including 3D production – have struggled to find extensive and reliable data to form the foundations of new generative AI tools.”
Lil Snack & GamesBeat
GamesBeat is excited to partner with Lil Snack to have customized games just for our audience! We know as gamers ourselves, this is an exciting way to engage through play with the GamesBeat content you have already come to love. Start playing games now!
Nunez added, “The path to get there involves making training dataset creation easy and highly scalable. GenAI 3D will require vast amounts of quality multi-camera datasets, specialized algorithms, computational techniques, and infrastructure to handle the additional dimensions effectively. Scale is crucial to developing generative AI models that possess a broad understanding of the incredible diversity in humans and other animals – and it’s that scale that Evercoast and Depthkit are best equipped to provide.”
Comprehensive 3D datasets haven’t existed due to the inherent complexity and expense of creating them, and consequently the algorithms have lagged behind, Evercoast said.
AI platforms from OpenAI, Meta, Google, Microsoft, Anthropic, and a growing list of others have been built on massive sources of text and 2D datasets, not 3D spatial data. Evercoast and Depthkit fill this void by bringing together the largest customer base for 3D volumetric video in the world, the companies said. Combined, the companies are uniquely able to deliver a highly scalable software platform capable of capturing and generating thousands of assets in a single day.
This capability is several orders of magnitude faster and 20 times less expensive than multi-camera arrays using traditional processing pipelines and expensive hardware, Evercoast said.
“Teaming up with Evercoast represents a thrilling new chapter for Depthkit and the entire 3D content creation community,” said James George, CEO of Scatter, the company that developed Depthkit, in a statement. “By joining forces, we are not just merging two leading technologies but also uniting our visions to create tools that are both innovative and accessible. We are confident our combined efforts will push the boundaries of spatial video and create new opportunities around the globe.”
Spatial video datasets and models tackle unique characteristics that 2D video struggles to effectively emulate at scale. These include real-world variability. Human motion is highly variable and influenced by numerous factors such as lighting conditions, occlusions, and diverse environments. Real-world datasets with multiple camera views provide a more comprehensive and accurate representation of these variations.
The spatial video datasets and modles also need depth and 3D understanding. Multi-camera setups enable the capture of more accurate depth information, which is essential for understanding the spatial relationships between body parts and clothing.
And training models using data from multiple camera views helps them generalize more effectively than monocular or synthetic datasets, particularly in scenarios with complex interactions where human performance involves other objects or individuals.
The growing base of specialized AI development in human performance capture, medical training, industrial simulation, robotics, sports performance, and higher education represents a massive addressable market for this technology.
VFX and other creative technologists also are using volumetric data combined with machine learning and AI video. These verticals tightly overlap between the two companies, and the combined technology stack further expands the market by delivering a hardware agnostic offering.
Moving forward, existing Depthkit customers will have full access to their existing features as well as a host of additional functionality through Evercoast. To learn more about Evercoast’s solutions for spatial video production, visit evercoast.com.
Evercoast is used by customers capturing celebrities like Peyton Manning, Drake and Charles Barkley in photo-realistic detail, as well as healthcare providers, universities, and broadcast media. Evercoast’s software enables datasets to be created anywhere, as evidenced by their clandestine trip into Ukraine at the outset of the war to record President Zelenskyy in 3D for global AR distribution.
Depthkit is a portable solution for creating high-quality volumetric video captures and livestreams using off-the-shelf depth sensors. Depthkit captures subjects in full motion, producing realistic human hologram performances to be placed in interactive 3D worlds. Volumetric video with Depthkit offers a revolutionary and affordable approach to 3D character creation that replaces traditional animation.
Read original article here
