Run AI models in your game
Sentis runs AI models on players’ devices for interactive gameplay.
Unity 6.0
Sentis 2.1 documentation LTS · Long-term supportSupported now
What changed
Unity feature: Sentis runtime AI
Sentis imports trained models and runs inference on the player’s CPU or GPU. It can turn model output into gameplay behavior without sending each inference request to a remote server.
Use it
- Install Sentis, import a supported ONNX model, and prepare its expected input tensors.
- Create an execution worker, run the model, and read its outputs in the game code.
Requirements and limits
- Sentis runs trained models; it does not train them. Supported operators, model size, and device resources determine whether a model is suitable.