Badges
Forums
REA ENGINE
User Guide: How to Activate Your AI Avatar
Follow these two phases to test the engine and fully activate the avatar s high-speed intelligence.
Phase 1: Vercel Test Run & Engine Check
Before connecting the AI, we provide a live environment to test the 3D performance and hardware compatibility.
Connection Link: avatar-projem.vercel.app
What to Check: Once you open the link, the system will initialize the 3D world. If you see the avatar and background moving smoothly, your 3D Engine and FPS (Bottom left) are working perfectly.
The Goal: At this stage, there is no AI connection yet. This is purely to verify the 1.4ms system latency and visual rendering.
Phase 2: Connecting the Gemini AI Brain
Once the engine is confirmed, follow these steps inside the app to give the avatar its "intelligence":
Open the Console: Click the "Engineering Console" (Connection Icon) on the right-side menu.
Input Your Credentials: In the Connection & AI tab, fill in the following fields exactly:
Provider: Google Gemini (Cloud)
API Key: Paste your personal Gemini API Key. (Get one for free at Google AI Studio)
Model Name: gemini-2.5-flash
TTS Model: gemini-2.5-flash-preview-tts
Verify: Click the test buttons at the bottom to seal the connection.
Results: Gemini Live & Real-Time Chat Active!
With the connection established, the system automatically unlocks its most advanced features:
Native Text Chat: Interact with the avatar via the text bar for instant replies.
Gemini Live (Voice Chat): The voice interaction becomes live. Thanks to Gemini Live technology, the avatar will listen to you in real-time, process thoughts instantly, and respond with perfect Native Lip-sync and expressions.
(Maker's Comment)
While developing this, one of our biggest challenges was running completely different skeleton structures (CC3 and RPM) in the exact same scene without corrupting each other's states (no more zombie bugs!). Thanks to our context-driven architecture, we solved this at the root, resulting in an incredibly flexible engine.
Your feedback, critiques, and questions mean the world to us. We are here all day to answer anything about the engine's architecture, the IK system, or our lip-sync algorithms!
Thanks for your support!
rea avatar engine
"Can we really achieve 0.8ms latency in the browser? Let's discuss Web-Native rendering limits."