Skip to main content

voxtral_micro/tts/
backbone.rs

1//! TTS backbone data structures (Q4 implementation in gguf/tts_model.rs).
2
3/// A single generated audio frame from autoregressive decoding.
4#[derive(Debug, Clone)]
5pub struct GeneratedFrame {
6    /// Semantic VQ index (raw, 0..8191 — without the +2 offset).
7    pub semantic_idx: usize,
8    /// Acoustic FSQ level indices per codebook (36 values, each in 0..20).
9    pub acoustic_levels: [usize; 36],
10}