Meshcapade

Meshcapade also offers foundation models that let digital humans perceive, understand and move around. The underlying technology is based on scientific knowledge of the human body and is presented through patented models. These models can be used to generate photorealistic digital avatars that can be created from text prompts, video or images.

Meshcapade's features include:

• Motion from Text: Type out any motion and use it immediately.
• Motion from Video: Capture motion with a single camera, no tripod needed.
• Body Shape from Images: Create avatars immediately from a single image, regardless of pose.

The company uses a generative model called SMPL (Skinned Multi-Person Linear), a whole-family 3D body model trained on hundreds of thousands of detailed 3D scans. SMPL captures the statistics of 3D body shape across the population, which means it's very accurate and realistic. It works with every game engine and graphics package because it's designed to.

Meshcapade hopes to offer an avatar-as-a-service interface that lets customers use realistic digital human avatars for e-commerce, the metaverse and other uses. With machine learning and computer vision, the company's platform can generate avatars from limited input data like photos, videos or body measurements.

The SMPL model is designed to be portable and to work with deep learning frameworks, which is why it's the de facto standard for 3D body models. It's available in female, male and non-binary versions and can be easily dropped into projects.

Meshcapade offers detailed API documentation and plugins to ease integration. The service is geared for companies that want to build services with realistic digital avatars, like games, social media and clothing companies.

Pricing isn't disclosed, but you can try the APIs directly in your browser, and there's a free trial option.

Published on June 14, 2024