HarmonyOS NEXT 什么时候可以适配虚幻引擎呢?
Looking really cool. Love the moment the suit activates. I guess this one is inspired from Ironman. And the moment you fly fast away reminds me a lot at the moment the enterprise activates warp ![]()
Thank You @Thorsten09 This was my first test using Metahuman thanks to @EmJay_1995 who design the outfit.
If I could solicit a few more devs to assist with this project I would revive it.
Are you still looking for people? Im available, ive been an unreal dev for over a decade
Yes Indeed.
The ATHENA - Gemini-UE5 Content Conductor is a conceptual visualization tool designed to illustrate how a Large Language Model (like Gemini) can be integrated deeply into a game engine (Unreal Engine 5) to act as a real-time “Game Master” or “Director.”
Instead of simply generating text or images offline, ATHENA runs alongside the game, manipulating the 3D world, logic, and narrative dynamically while the player is playing.
Here is a breakdown of the system based on the visualizer:
1. The Three-Panel Architecture
The interface is divided into three distinct sections representing the flow of data:
- Left Panel: UE5 Runtime World (The Client)
- What it represents: This is the actual game running on the player’s screen.
- Visuals: It renders “Solid” geometry and entities. It simulates the high-performance physics and rendering thread (120 FPS).
- Role: It executes the commands. If ATHENA says “Spawn Enemy,” this panel is where the 3D model actually appears and starts moving.
- Center Panel: Sync Bridge (The Pipeline)
- What it represents: The networking layer (WebSockets/HTTP) connecting the game engine to the AI.
- Visuals: A “Sync Meter” that fluctuates, representing latency and data throughput.
- Role: It ensures the AI’s understanding of the world matches the actual game state. If connection is lost, the AI cannot direct the game effectively.
- Right Panel: ATHENA Spatial Mirror (The Brain)
- What it represents: Gemini’s internal “mental model” of the game world.
- Visuals: It renders the same world but as Wireframes and Data Points. It visualizes pathfinding vectors (dotted lines) and entity IDs rather than graphics.
- Role: It processes logic. It calculates where an enemy should spawn or what a character should say based on the spatial context, then sends a JSON directive back to the Left Panel.
2. The “Director” Controls (Footer)
The buttons at the bottom simulate the specific API Commands that ATHENA can issue to the game engine:
- Gen Architecture: Spawns physical geometry (walls/cover) dynamically, perhaps to block a player or create a bridge.
- Gen Objectives: Drops logic nodes (like a “Capture Zone”) into the 3D space, effectively creating a new game mode on the fly.
- Gen Entity: Spawns AI agents with specific behaviors (e.g., “Stalker” class) and autonomous navigation logic.
- Direct Narrative: Injects dialogue lines into specific characters based on context (e.g., “Movement detected in Sector 4”).
- Direct Cutscene: Takes control of the game camera, adding cinematic black bars and changing the field of view for dramatic effect.
- Direct Audio: Swaps the background music track or soundscape intensity to match the current tension level.
3. Data Flow Simulation
When you click a button, the visualizer simulates the Bi-Directional Communication:
- Request: A stream of light shoots from the UE5 panel to the Gemini panel (sending the query).
- Processing: The Gemini panel logs the thought process (e.g., “Analyzing Spatial Context…”).
- Directive: A JSON payload (e.g.,
{ type: 'NPC_ENEMY', class: 'STALKER' }) is generated. - Execution: A stream of light shoots back to the UE5 panel, and the new content appears in the game world.




