Human AI

Overview

Human AI lets users create and converse with custom 3D avatars powered by OpenAI's latest models. Both a consumer product and an enterprise integration kit, it ships pre-built avatar presets and an API for embedding personalities into other applications.

We were brought in to design and engineer the platform end-to-end — from real-time 3D rendering and voice pipeline to OpenAI orchestration and billing.

What we built

• AvatarsReal-time 3D avatar engine with lip-sync, gaze, and expression driven by streaming model output.
• VoiceLow-latency voice in/out pipeline (STT → LLM → TTS) with interruption handling.
• PresetsLibrary of curated personas with distinct voices, behaviors, and visual styles.
• APISDK for embedding Human AI avatars into third-party web and mobile apps.
• AdminUsage analytics, role management, and billing integrated with Stripe.

Architecture

WebGL rendering on the client (Three.js) with model state synchronized over WebSocket. OpenAI orchestration runs server-side on Node.js, with caching and rate-limit handling at the edge. Audio is streamed in chunks for sub-second response.

Outcome

Shipped at human-ai.app, with consumer and B2B integration pilots across Australia and Asia-Pacific.