How to Quickly Build 3D Virtual Avatars with Front‑End Techniques
This article explains the end‑to‑end front‑end workflow for creating real‑time 3D virtual avatars using the Oasis engine, covering glTF asset handling, model and texture replacement, bone animation mapping, inverse kinematics, physics integration, AI enhancements, and a low‑code development approach.
Overall Process of Virtual Avatar
Artists model the avatar in a 3D tool, but front‑end developers must intervene early to ensure mobile rendering constraints such as triangle count and to define naming conventions for bones, skins, and materials. The artist exports a glTF file, which becomes the basis for all front‑end work.
Oasis Engine Core Capabilities
The Oasis engine provides fundamental features for virtual humans, including morph and skinning animation, material rendering, and a resource loader that easily imports glTF files and decomposes them into components and entities.
Asset Replacement Techniques
Model Replacement
Fixed‑shape assets like eyes or equipment are placed under a designated
Headnode in the glTF hierarchy, allowing artists to model them relative to the head origin.
Texture Replacement
Non‑geometric parts such as beards or blush are swapped by overlaying new texture maps onto predefined regions of the facial texture.
Skin Replacement
Skins are linked to multiple bones; by adhering to naming conventions, developers can locate and replace skin meshes programmatically.
Bone Animation Mapping
To avoid distortion when applying a single animation set to avatars of varying body shapes, bone mapping discards translation and scale data, preserving only rotation so the animation fits any proportion.
Inverse Kinematics and Physics
Inverse kinematics computes joint poses from end‑effector targets (e.g., feet, hands) to adapt avatars to ground contact or focus points. Simple cases use analytical solutions; complex rigs employ iterative methods like FBIK.
Collision handling uses PhysX.js compiled to WebAssembly for basic collision detection, and Nvcloth.js for cloth simulation to prevent mesh interpenetration.
AI End‑to‑End Capabilities
AI technologies such as text‑to‑speech drive facial and skeletal animation, while motion capture data is remapped via bone mapping to avoid distortion on avatars with different body proportions.
Low‑Code Virtual Avatar Development
All aforementioned functionalities are encapsulated in the Ark SDK, which integrates with the Oasis editor plugin. Artists upload assets, perform face sculpting, scene assembly, material tuning, and animation sequencing, then export front‑end components, allowing developers to focus on business logic.
Complex animation logic is managed through an editor‑based animation orchestration UI, where animation layers and conditional scripts define when and how actions are triggered.
Future Outlook
With the rise of the metaverse and Web3, virtual human production cost reduction remains critical. Ongoing research targets animation, rendering, physics, and AR/VR integration to place avatars in real environments. The Oasis editor and Ark components will be open‑sourced early next year.
Alipay Experience Technology
Exploring ultimate user experience and best engineering practices
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.