move.ai

In recent years, motion capture technology has revolutionized the film, gaming, and animation industries, enabling the creation of increasingly realistic digital characters and environments. While traditional marker-based systems have dominated the market, new developments in markerless motion capture are beginning to transform how we capture and animate human movement. At the forefront of this technological shift is Move.ai, a company that has developed groundbreaking AI-powered solutions for extracting natural human motion from video. In this article, we explore the technical capabilities, applications, and implementation requirements of Move.ai's markerless motion capture technology, examining how it is being used to push the boundaries of digital animation across multiple industries.

Company Overview

Move AI's technology combines AI, computer vision, and biomechanics to extract natural human motion from video. The company developed the app with seed funding of $10M and has established itself in the industry through precise hand and finger tracking capabilities.

The technology enables high-quality motion data capture in various environments using standard phones and cameras. For detailed setup, users need to configure camera layouts based on the space size, using supported FLIR cameras with appropriate lens configurations. The system requires a minimum of 60fps camera operation and 10GB network throughput for optimal performance.

Move AI offers three main product plans to cater to different needs:

• Move One: A simple, affordable motion capture solution from a single camera, available on web, iPhone, and iPad. It tracks a single person in up to 5m x 5m volumes at 60 FPS.

• Move Live: A real-time motion capture system that tracks up to 4 people in volumes up to 14m x 14m. It captures up to 110 FPS with 4-8 cameras and supports both solo operation and full autonomy.

• Move Pro: The most comprehensive option, tracking up to 22 people in volumes up to 100m x 50m. It captures at 120 FPS with support for up to 12 cameras, including GoPro and Sony models.

The company has established itself in diverse industries through its markerless motion capture solutions. Clients include major studios like Electronic Arts, where the animation director praised the technology, and Eye Garden, which collaborated on the MGMT music video "Mother Nature."

Technical Specifications and Operations

Move AI's platform captures up to 22 people simultaneously across volumes as large as 100m x 50m, while its Move Live system handles up to 4 people in spaces up to 14m x 14m. The technology processes data through local neural networks to generate accurate 3D representations while maintaining low latency.

The system utilizes deep learning models to detect key points within 2D images from multiple cameras, with the ability to track hand and finger movements precisely. For optimal performance, the platform requires cameras running at 60fps, with setups supporting 4 to 8 cameras depending on the deployment plan. Users configure camera layouts based on the space size, selecting appropriate lens configurations that vary from 2.8mm to 4mm focal lengths.

The complete setup process typically takes less than an hour, including calibration, which can be completed in just one minute. The system produces latencies of approximately 100ms, dependent on network configuration, while transmitting data through standard web interfaces compatible with both solo operation and full autonomous workflows.

For data processing, the platform employs local neural networks to reconstruct 3D motion from 2D inputs, achieving frame rates up to 110fps for the Live system and 120fps for professional deployments. The software supports multiple export formats including FBX and USD, allowing for easy integration with popular 3D animation applications while maintaining the ability to export raw motion data for post-production use.

Use Cases and Applications

From animating Pokémon trainers to assisting major studios like Electronic Arts, Move AI has established itself in diverse industries through precise hand and finger tracking capabilities. The technology supports both real-time applications and high-fidelity post-production through FBX and USD exports, making it a versatile tool for 3D animation workflows across various industries.

The company's approach to motion capture stands out through its markerless technology, enabling high-quality data extraction in any environment without the need for suits or specialized studios. This capability has proven particularly valuable for creators looking to integrate physical performance with digital animation, as demonstrated by animator Daniel Contreras, who uses Move AI to bring characters to life while maintaining creative control over his projects.

Move AI's technology has made significant strides in real-time applications, as seen in its collaboration with XiteLabs for Grimes' XR live performances and its use by Electronic Arts for game development. The platform's ability to capture up to 22 people simultaneously across large volumes of up to 100m x 50m has facilitated complex multi-person animations, while its smallest deployment option, Move One, provides an affordable solution for capturing a single person in 5m x 5m spaces.

The company has demonstrated its technology's versatility through various deployments across different industries and project sizes. Notable applications include:

• Creating a digital twin soccer match in the SOIL metaverse platform using data from a real-life match at Wembley Stadium

• Enhancing Nike's Euro 2024 activations in store environments across multiple cities

• Accelerating Ubisoft's 3D animation workflow for Just Dance 2023

• Assisting Eye Garden on MGMT's "Mother Nature" music video

Through these diverse applications, Move AI continues to push the boundaries of 3D animation, demonstrating the potential of markerless motion capture to transform how we create and experience digital content.

System Requirements and Setup

The system configuration process requires careful planning to meet the specific needs of the deployment environment. A minimum of 60fps camera operation is essential for optimal performance, with camera layouts optimized for different space sizes ranging from small (5m x 5m) to large (100m x 50m) volumes.

The company supports a specific FLIR camera model, the BFS-PGE-16S2C-CS, with three recommended focal lengths: 2.8mm, 3.5mm, and 4mm. Camera placement is critical, with optimal positioning typically achieved when cameras are mounted directly above the capture volume, though the recommended height limit is 3m to maintain image quality and field of view.

For larger deployments, up to 12 cameras can be used, supporting various camera types including GoPro and Sony models. The server requirements are substantial, needing a CPU comparable to an Intel Core i9-12900K or AMD Ryzen Threadripper PRO 7965WX, along with an NVIDIA RTX A6000 GPU and at least 32GB of DDR4 RAM. The system architecture includes specific networking requirements: a 2000W power supply, 1TB NVMe SSD storage, and a dedicated 10Gb/s network interface card.

The setup process begins with camera configuration, which typically takes less than an hour to complete, including both hardware placement and initial software calibration. Intrinsic camera calibration requires running at 60fps and can be performed using either a chessboard pattern or a custom board with known dimensions. Extrinsic calibration, which captures relationships between camera positions, follows the same process but requires multiple camera setups.

The software provides real-time feedback during capture through an on-screen status indicator, showing elapsed recording time and frame loss. Post-capture processing allows for basic retargeting and export capabilities, supporting FBX, USD, and MP4 file formats. The entire system, including technical account management for larger deployments, operates through a web-based interface that manages both single and multi-camera setups efficiently.

Pricing and Payment Models

Move AI offers flexible pricing models tailored to different deployment needs, from single-camera setups to enterprise-level productions.

Basic Plans and Usage

The company's most accessible offering is Move One, a single-camera motion capture app available on iOS devices that captures HD video at 60fps with up to 60-second recording limits per session. The app allows hand and finger tracking while supporting basic 3D export formats including FBX, USD, and MP4.

For larger or more frequent projects, the company offers three tiered subscription plans:

• Starter: $15 per month including 180 credits, suitable for infrequent users

• Basic: $35 per month with 480 credits, ideal for regular use

• Team: $75 per user per month with a minimum of three users and a maximum of ten, featuring 4,000 shared credits

Each credit represents one second of animation for a single person, with credits expiring at the end of each month for paid plans and never expiring for free accounts.

Customized Solutions and Large Deployments

For more complex projects or enterprise-level deployments, Move AI offers customized plans supporting multiple camera systems and advanced features. The company's smallest multi-camera option, Move Live, tracks up to four people in volumes up to 14 meters by 14 meters using between four and eight cameras. This system captures at 110fps with approximately 100ms latency and requires one-minute calibration before each session.

The company's premium offering, Move Pro, enables capture of up to 22 people across volumes up to 100 meters by 50 meters, supporting up to 12 cameras including models from GoPro and Sony. This system captures at 120fps with advanced export capabilities including 4K resolution and support for FBX, USD, and MP4 file formats.

Both Multi-Cam systems require significant upfront investment in hardware infrastructure, including server-grade computing, specialized cameras, and precise calibration equipment. The company provides detailed technical specifications and recommended configurations to help customers design appropriate deployment environments, though specific implementation details vary based on project requirements and space dimensions.