Flexible AI model implementation
Sentis allows for standardized AI model implementations such as style transfer and speech recognition. Models can be customized through model weights, layers, and chaining multiple inferences.
High performance on user devices
Sentis leverages the compute power of end-user devices rather than the cloud, which eliminates complex cloud infrastructure, network latency, and recurring inference costs.
Use cases
Sentis is a framework for importing and running third-party AI models but does not include its own models. Explore models on Hugging Face and find more here.
Object identification
Detect, classify, and segment objects with an in-game or on-device camera.
Customizable AI opponents
Power a board game opponent with specific rules and custom difficulty curves. Run a neural network trained on the game rules and determine game win probabilities after each move.
Handwriting detection
Identify handwritten numbers, letters, and symbols for unique gameplay interactions.
Depth estimation
Estimate the depth of real-world objects in an augmented reality view to occlude objects in a game scene.
Speech recognition
Convert live speech to in-game text using a machine learning language model for natural language interactions between players.
Smart NPCs
Automate dialogue and create meaningful interactions between players and NPCs without the limitations of manual scripting.
Resources
AI-driven gameplay
Learn how running neural network models on end-user devices can activate dynamic, living worlds like never before.
Technical samples
Explore how to enable simple features using Unity Sentis with samples for digit recognition, depth estimation, and creating reactive AI NPCs.
Technical documentation
Dive in with documentation and samples for Unity Sentis, and power up your game with AI.
Community discussion
Join a community of developers building the next generation of AI-powered games and applications.