Expand description
Enhanced optimizations for whisper-cpp-plus
This module provides performance optimizations inspired by faster-whisper while maintaining compatibility with the base whisper.cpp API.
§Features
- Enhanced VAD: Intelligent speech segment aggregation for optimal chunk sizes (preprocessing)
- Temperature Fallback: Quality-based retry mechanism for difficult audio (transcription)
- Performance: 2-3x speedup on audio with silence, improved accuracy on noisy audio
§Architecture
The enhancements are designed as orthogonal improvements:
- VAD enhancement is a preprocessing step that happens BEFORE transcription
- Temperature fallback is a transcription enhancement for quality
- Both can be used independently or together
Re-exports§
pub use vad::EnhancedWhisperVadProcessor;pub use vad::EnhancedVadParams;pub use vad::EnhancedVadParamsBuilder;pub use vad::AudioChunk;pub use fallback::EnhancedTranscriptionParams;pub use fallback::EnhancedTranscriptionParamsBuilder;pub use fallback::QualityThresholds;pub use fallback::EnhancedWhisperState;pub use fallback::TranscriptionAttempt;