Description
NVIDIA ACE for Games is a suite of AI tools for middleware and game developers building knowledgeable, actionable, and conversational in-game characters, in NVIDIA's own description. Rather than a single app, it's a set of cloud and on-device AI models and SDKs covering speech, character intelligence, and animation, designed to be integrated into existing game engines and pipelines.
Read More
The stack includes small AI models optimized for on-device inference on gaming hardware, tuned for accuracy and low latency within a small memory footprint, plus the NVIDIA In-Game Inferencing (NVIGI) plugins that schedule AI inference alongside a game's graphics workload. Model options include NVIDIA Nemotron for automatic speech recognition, supporting up to 25 languages, and small language models for game-agent reasoning, plus third-party text-to-speech models such as Resemble AI's Chatterbox and Alibaba's Qwen 3 TTS, both offering zero-shot voice cloning.
For Unreal Engine 5, NVIDIA ships ready-made plugins under the MIT license: Audio2Face-3D for facial animation, with a matching Autodesk Maya ACE plugin for offline animation, an ASR plugin for speech recognition, an LLM plugin for local text generation and function calling, and a TTS plugin for voice output. For custom engines, the Game Agent SDK (in beta) is a lightweight C/C++ framework with Agent, Chat, and RAG APIs to connect NPCs to game state and knowledge, alongside the In-Game Inferencing SDK for running models on-device via CUDA.
NVIDIA cites several shipped examples: Total War: PHARAOH's AI advisor, Streamlabs and Inworld AI's streaming sidekick, KRAFTON's PUBG Co-Player Characters, Wemade Next's MIR5 adaptive AI bosses, and KRAFTON's inZOI Smart Zois. ACE for Games is a developer platform, not a consumer product: it's aimed at game studios and technical artists building NPC and AI-companion systems inside Unreal Engine or custom engines, with SDKs, plugins, and models available to download directly from NVIDIA's developer site.
- CategoryAI Game Development Tools
- Pricing ModelFree
- DeploymentCloud, On-Device
- PlatformsUnreal Engine
NVIDIA ACE for Games's Core Features
Small, on-device AI models optimized for gaming hardware, covering speech, language, and vision
NVIDIA In-Game Inferencing (NVIGI) plugins that schedule AI inference alongside graphics workloads
NVIDIA Nemotron ASR: automatic speech recognition supporting up to 25 languages
Local LLM plugin for low-latency character dialogue and function calling
Audio2Face-3D facial animation plugin for Unreal Engine 5 and Autodesk Maya
Text-to-speech models with zero-shot voice cloning, including Chatterbox and Qwen 3 TTS
Game Agent SDK (beta): a C/C++ agentic framework with Agent, Chat, and RAG APIs
Ready-made Unreal Engine 5 plugins under the MIT license for ASR, LLM, TTS, and animation
Getting Started with NVIDIA ACE for Games
Download the SDK: get the NVIDIA ACE Game Agent SDK or In-Game Inferencing SDK from the developer site.
Pick your models: choose on-device or cloud models for speech, intelligence, and animation.
Integrate with your engine: use the Unreal Engine 5 plugins, or the C/C++ SDK for a custom engine.
Connect character logic: use the Agent, Chat, and RAG APIs to link NPCs to game state and knowledge.
Animate and voice the character: apply Audio2Face-3D for facial animation and a TTS plugin for speech output.
NVIDIA ACE for Games's Use Cases
- AI NPC dialogue
- Facial animation from audio
- Adaptive AI companions
- Voice for game characters
