LLMUnity

Why we included this project

Unity developers who want NPCs that actually talk back will find a lot to like here. LLMUnity plugs a large language model straight into the engine: attach a character component and start a dialogue with one line of code, and the model runs locally on CPU or GPU, whether the target is PC, mobile, or VR. A bundled RAG system lets each character answer from your own lore or scripted data instead of improvising, so the NPC sticks to what it knows. Because inference happens on-device with no internet round-trip, dialogue stays responsive during gameplay and player data never leaves the game, which matters for both latency and privacy. It ships with a llama.cpp-based backend that supports all major open models, plus example scenes for chat, voice, and animated characters, making it a practical starting point for shipping a cast of AI-driven characters rather than just a prototype.