Spatius

Spatius is a real-time AI avatar infrastructure platform that shifts visual rendering from the cloud to the user's device. By transmitting lightweight motion data instead of streaming compressed video, the platform enables developers to deploy interactive 3D avatars across web and mobile platforms with minimal latency and bandwidth consumption. Developers integrate Spatius directly into their existing voice AI pipelines (ASR, LLM, and TTS) to handle the visual rendering layer.

Key Features:

On-Device 3D Gaussian Splatting (3DGS): Compiles 3D avatar assets from a single photo into a compact 5–10 MB model executed locally on client hardware. This mechanism delivers 60 fps avatar rendering on entry-level chipsets without requiring a dedicated GPU, reaching 99% of active iOS, Android, and Web devices.

Low-Bandwidth Motion Server: Converts input text-to-speech (TTS) audio into lightweight lip-sync and body motion data (10–20 KB/s) in the cloud. This reduces data transmission by approximately 99% compared to traditional cloud-based video streaming (1–2 MB/s), lowering network costs and improving performance on constrained connections.

Multi-Platform Client SDKs (AvatarKit): Compiles native rendering runtimes for WebGL/WebGPU (Web), Metal (iOS), Vulkan (Android), and Flutter. The SDK downloads avatar assets, executes local rendering, and synchronizes the visual lip and body movement with local audio playback.

Modular Pipeline Integration: Isolates the rendering layer from the conversational logic, accepting standard TTS audio inputs to drive motion generation. This allows developers to plug in their own Automated Speech Recognition (ASR), Large Language Model (LLM), and Text-to-Speech (TTS) components.