About
Unity Sentis is a powerful AI inference engine built into the Unity game development platform, enabling developers to run trained neural network models (in ONNX format) directly at runtime inside games and interactive 3D applications. As part of the broader Unity AI suite, Sentis allows game studios to integrate machine learning capabilities—such as character behavior, image recognition, procedural generation, and real-time decision-making—without relying on external servers or cloud dependencies. Sentis is deeply embedded in the Unity Editor workflow, offering contextual awareness so AI assistance is tailored to your specific project assets and structure. It supports curated third-party AI models that can be loaded and used without manual setup or switching between tools. Privacy-conscious teams will appreciate the platform's transparent data controls, which feature no default data collection or model training on your project content. With Unity Sentis, developers can leverage pre-trained models for a wide range of in-game AI tasks, from NPC pathfinding and enemy behavior to style transfer and object detection. The library is platform-agnostic, matching Unity's own 25+ deployment targets including mobile, PC, console, and XR. It's suited for indie developers, AAA studios, and enterprise teams building immersive training simulations or interactive experiences. Unity Sentis lowers the technical barrier for bringing ML-powered features into real-time 3D applications, all from within the familiar Unity environment.
Key Features
- ONNX Model Inference at Runtime: Load and execute standard ONNX-format neural network models directly within a running Unity game, no external server or cloud connection required.
- In-Editor Contextual AI Assistance: Unity AI provides tailored, context-aware help by referencing your actual project assets and Editor state, making suggestions relevant to what you're building.
- Curated Third-Party Model Support: Access and use pre-vetted third-party AI models inside Unity without manual configuration or switching to external applications.
- Transparent Data Controls: Full transparency on data usage with no default data collection or training on your project content, giving studios full control over their IP.
- Cross-Platform Deployment: Deploy AI-powered Unity projects across 25+ platforms including mobile, PC, console, and XR without re-engineering inference pipelines for each target.
Use Cases
- Implementing intelligent NPC behavior using trained reinforcement learning models that run locally within a game build.
- Applying real-time style transfer or visual effects powered by neural networks without cloud round-trips.
- Building gesture or pose recognition for XR games using computer vision models executed on-device via Sentis.
- Prototyping and iterating on ML-powered game mechanics directly inside the Unity Editor with contextual AI assistance.
- Deploying AI-driven features across mobile, console, and PC from a single Unity project without platform-specific ML infrastructure.
Pros
- No External Infrastructure Needed: AI model inference runs entirely on-device, eliminating cloud latency, costs, and connectivity dependencies for real-time game applications.
- Seamless Unity Editor Integration: Sentis fits naturally into existing Unity workflows, reducing the learning curve for developers already familiar with the Unity ecosystem.
- Privacy-First Design: No default data collection or model training means studios retain full ownership of their game assets and project data.
- Broad Platform Compatibility: Leverages Unity's extensive platform support, so ML features can ship on mobile, console, PC, and XR from a single codebase.
Cons
- Limited to ONNX Model Format: Sentis is currently built around the ONNX standard, which may require additional model conversion steps for teams using frameworks like TensorFlow natively.
- Still in Beta / Evolving Feature Set: As an evolving Unity AI product, some features and supported operators may be incomplete compared to dedicated ML inference runtimes like TensorRT or Core ML.
- Requires Unity Editor: Sentis is tightly coupled to the Unity development environment, making it inaccessible to developers using other game engines or standalone ML pipelines.
Frequently Asked Questions
What is Unity Sentis?
Unity Sentis is Unity's AI runtime inference library that allows developers to embed and run ONNX neural network models directly inside Unity games and real-time 3D applications, with no cloud or external server required.
Is Unity Sentis free to use?
Yes, Unity Sentis is available as a free package within the Unity ecosystem, accessible to all Unity plan tiers including personal and indie plans.
What model formats does Unity Sentis support?
Unity Sentis supports the ONNX (Open Neural Network Exchange) format, which is compatible with models trained in PyTorch, TensorFlow, and other popular ML frameworks after export.
Does Unity Sentis require an internet connection at runtime?
No. Sentis runs inference entirely on-device, meaning your game's AI features work without any internet connection or cloud API calls during gameplay.
What kinds of AI features can I build with Unity Sentis?
You can build features like intelligent NPC behavior, object and gesture recognition, style transfer, procedural animation, real-time image classification, and any other capability expressible as a trained neural network model.
