zh | eng


3D Loomie™ avatar platform for enterprise applications

Powered by deep learning and advanced computer graphics algorithms, Loom.ai provides the most robust and scalable 3D avatar platform for enterprise partners and developers. Loom.ai's SDKs and editor tools are available for native integration on mobile (Android, iOS) and PC (linux, OSX, Windows), or accessible on the cloud via a REST API. Loom.ai also provides API wrappings in Python or Go for deployment on partners' cloud backend.

Avatar SDK

Create a Loomie

An SDK for building a personalized Loomie from a single front-facing photograph of a user's face using deep learning. The key features include:

    •  Solve time: less than 5 secs on mobile CPU

    •  Texture: photo pixels or stylized
    •  Facial rig: FACS-style controls for face, gaze, tongue.

    •  Optional: Facial editor with manual attribute controls
    •  Optional: fully body rig with FK/IK
    •  Optional: OpenGL based renderer SDK for mobile/PC.

Editor SDK

Loomie customization

An SDK to customize the 3D accessories that attach to a Loomie face or body. Features include:

    •  Accessory fitting: real-time SDK for mobile and PC
    •  Fitted elements: hair, glasses, beard, head gear, etc
    •  Additional fitting: body type, outfits
    •  Animation: Keyframe or tracking, face, body & gestures
    •  Content: Maya-based pipeline with cloud storage

Video2Animation SDK

Animate from Video

An SDK for animating a FACS rig in real-time from RGB video and audio using deep learning. Features include:


    • Face detection: Mobile and PC versions

    • Face tracking: Highest quality in market

    • Specs: Facial expressions and lip sync at 60fps

    • Platforms: Mobile (Android, iOS) and PC

    • Output: Real-time animation curves on FACS rig

    • Optional: Live streaming to Unity/UE4

Voice2Animation SDK

Animate from Voice

End-to-end real-time SDK for automatic lip-sync animation from audio. Features include:


    • Generalizes to different languages and people

    • Very low latency, ideal for real-time applications

    • High inference speeds (500fps+)

    • Optimized for ARM/Neon and x86 architectures

    • Additional controls for emotion triggering and gaze

    • Small model size, ideal for client integration