vad_plus 0.3.1

Silero VAD ONNX voice activity detection FFI plugin for Flutter.

0.3.1

• Re-publish of 0.3.0 (no functional changes).

0.3.0

• Fix pre-speech padding: segments now start with the full preSpeechPadFrames (~96 ms) of lead-in and no longer duplicate the triggering frame.
• Implement endSpeechPadFrames: emitted segments keep a short silence tail after the last voiced frame instead of the whole ~768 ms redemption window (previously the parameter had no effect).
• Fix per-event native memory leak on Android — event payloads are now freed on a short grace timer after delivery, mirroring the iOS/macOS cleanup.
• Fail with an error when a provided modelPath does not exist on Android (previously fell back to the bundled model silently).
• Re-extract the bundled model on every Android init so a plugin update is never shadowed by a stale cached copy.
• Release the native handle when initialize() fails, so a failed instance no longer leaks and can be retried.
• Align VADEvent in src/vad_plus.h with the actual native ABI (flat struct) and regenerate the FFI bindings — dart run ffigen is now safe to re-run.

0.2.1

• Complete README: quick start, events, configuration, VAD algorithm, processAudio mode, per-platform permissions, architecture, and known limitations.
• Restructure the example app into feature-based modules.

0.2.0

• Require Dart SDK ^3.9.0

0.1.2

• Update ffi dependency to ^2.2.0

0.1.1

• Code quality optimization

0.1.0

• Add support for 16KB memory page size on Android.

0.0.1

• Initial release.