2.0.6

• Now, all model classes (VadWebRTC, VadSilero, and VadYamnet) are closable.
• Updated JavaDoc for all files.
• Updated Build Gradle dependency.

2.0.5

• Updated logic for processing false positive results;
• Deprecated setContinuousSpeechListener() method;
• Added proper validation for speechDurationMs() and speechDurationMs();
• Updated Tensorflow Lite Audio library;
• Refactoring;
• Updated all dependencies.

2.0.4

• Added support of different audio data types;
• Updated dependencies and lowered minSdkVersion;
• Fixed issue related to initialization of maxSilenceFramesCount and maxSpeechFramesCount.

2.0.3

• Introduced a new model based on Deep Neural Networks (DNN) - Yamnet VAD, which provides audio classifier for 521 different type of sound;
• Split libraries to different modules;
• Fixed issue in WEBRTC VAD.

2.0.2

• WebRTC VAD has been updated and multiarch compilation has been properly configured.

2.0.1

• Introduced a new model based on Deep Neural Networks (DNN) - Silero VAD DNN, resulting in improved speech detection capabilities;
• Global Refactoring: The project underwent a comprehensive refactoring process, improving code structure, readability, and maintainability;
• Transition to Kotlin: The entire project has been migrated from Java to Kotlin, leveraging the benefits of Kotlin's concise syntax and enhanced features;
• Library and Dependency Updates: All libraries and dependencies have been updated to their latest versions, ensuring compatibility and incorporating bug fixes and performance improvements;
• Added Kotlin Bom which will properly align all dependencies;
• Integration of ONNX Runtime: The library now includes ONNX Runtime dependencies, enabling efficient execution of ONNX models and expanding the range of supported models.

1.0.1

• Upgraded all libraries;
• Added comments for every method in Vad.java;
• Removed andoridx.appcompat from VAD project;
• Fixed issue with wrong Frame Sizes in MainActivity.java.

1.0.0

• First release of Android VAD library.