//! Kokoro-82M model implementation.
//!
//! Architecture: PL-BERT → Text Encoder → Prosody Predictor → ISTFTNet Decoder → 24kHz audio
//! StyleTTS2-based approach with lightweight 82M parameters.
//!
//! Voice cloning is supported when the checkpoint includes the style encoder
//! weights (`style_encoder.*` and `predictor_encoder.*` prefixes).
pub use KokoroModel;