Kinetix

What it is

Kinetix is an AI research company that focuses on modeling, understanding, and generating human motion in three dimensions. The organization develops generative models and foundational technologies for 3D character animation, drawing on work in human pose estimation and computer vision to support end-to-end character workflows. Its product set includes a video-conditioned animation pipeline that can reconstruct animated 3D character motion from RGB acting footage and a controllable video generation model named Kamo-1, which uses 3D inputs to inform rendered shots. The company positions these capabilities for integration into creative tools and production pipelines, and its team of engineers and researchers works on deploying those models in practical applications.

Key features

Kinetix’s offerings emphasize multimodal conditioning and physical grounding. Kamo-1 accepts combinations of 3D camera trajectories, 3D character animations derived from RGB video, reference images, and natural language prompts to steer both camera and character motion within the same generated clip. The models leverage depth and pose estimates as control signals so generated frames align with spatial layout and articulated motion. The pipeline supports reconstructing scenes, placing a 3D character mesh, animating that mesh over time, and rendering a final video using a flow-matching generation process. Kinetix also provides a video-to-animation model packaged for integration, and the company has supplied moderation tooling for user-generated animation features.

Use cases

Kinetix’s technology is applied where controllable 3D motion and coordinated camera movement are required. Game developers can generate character animations through native integrations such as Unity Muse, and game teams have used the video-to-animation pipeline to let players create custom emotes in live titles, with moderation layers for safety. Creative platforms and middleware providers have licensed Kinetix models to add AI-driven animation features to services like Adobe Mixamo. Research groups and production studios can use the models for pose estimation tasks, reconstructing performance-driven animation from video, and producing AI-generated video that respects physical and articulated constraints.