Skip to main content

whisper_cpp_plus/enhanced/
mod.rs

1//! Enhanced optimizations for whisper-cpp-plus
2//!
3//! This module provides performance optimizations inspired by faster-whisper
4//! while maintaining compatibility with the base whisper.cpp API.
5//!
6//! ## Features
7//!
8//! - **Enhanced VAD**: Intelligent speech segment aggregation for optimal chunk sizes (preprocessing)
9//! - **Temperature Fallback**: Quality-based retry mechanism for difficult audio (transcription)
10//! - **Performance**: 2-3x speedup on audio with silence, improved accuracy on noisy audio
11//!
12//! ## Architecture
13//!
14//! The enhancements are designed as orthogonal improvements:
15//! - VAD enhancement is a preprocessing step that happens BEFORE transcription
16//! - Temperature fallback is a transcription enhancement for quality
17//! - Both can be used independently or together
18
19pub mod fallback;
20pub mod vad;
21
22pub use fallback::{
23    EnhancedTranscriptionParams, EnhancedTranscriptionParamsBuilder, EnhancedWhisperState,
24    QualityThresholds, TranscriptionAttempt,
25};
26pub use vad::{
27    AudioChunk, EnhancedVadParams, EnhancedVadParamsBuilder, EnhancedWhisperVadProcessor,
28};