com.unity.ai.inference 2.4.0
Source docsNotes: processedReleased October 30, 2025
Unity Compatibility
Minimum Unity: 6000.0
Package Dependencies
- com.unity.burst (1.8.17)
- com.unity.collections (2.4.3)
- com.unity.dt.app-ui (1.3.1)
- com.unity.modules.imageconversion (1.0.0)
- com.unity.nuget.newtonsoft-json (3.2.1)
✨ Features
- LiteRT model import
- Tokenization API
- STFT and DFT ONNX operators
- BlackmanWindow, HammingWindow, HannWindow and MelWeightMatrix ONNX operators
- BitwiseAnd, BitwiseOr, BitwiseXor, BitwiseNot ONNX operators and functional methods
- AsStrided, Atan2, Expm1, Log10, Log1p, Log2, Rsqrt, Trunc, ReduceVariance, Diagonal layers, functional methods and optimizer passes
- NotEqual, FloorDiv, TrueDiv layers and LiteRT operators
📈 Improvements
- Renamed Inference Engine to Sentis in package name and documentation
- Improved model import time for ONNX models
- ONNX model import operator order now consistent with the original model
- Improved optimization passes to reduce operator count in imported models
- Improved visualizer loading times and consistency in displaying attributes
- ScatterND operator can now run on much larger tensors, enabling new models
- ScatterND operator now allows negative indices
- ONNX model outputs that are not connected to any inputs are no longer incorrectly pruned
- Improve model import warning and error display in the inspector
🔧 Bug Fixes
- Small errors in documentation
- Faulty optimization passes that could lead to inference issues
- Memory leaks on model constants
- Non-matching ProfilerMarker calls
- Issues in CPU callback which could lead to incorrect inference on some models
- Enable missing modes for GridSample and Upsample operators