Module style 

Voice style transfer module (GH-132).

Provides voice style transfer primitives for:

• Prosody transfer (pitch, rhythm, energy)
• Timbre conversion (spectral characteristics)
• Cross-lingual style transfer

Architecture

Source Audio → Content Encoder → Linguistic Features
                                        ↓
Reference Audio → Style Encoder → Style Vector → Decoder → Styled Audio

Example

use aprender::voice::style::{StyleConfig, StyleVector, prosody_distance};

let style_a = StyleVector::new(vec![0.5, 0.3, 0.2], vec![0.1, 0.2, 0.3], vec![0.4, 0.5, 0.6]);
let style_b = StyleVector::new(vec![0.6, 0.4, 0.3], vec![0.2, 0.3, 0.4], vec![0.5, 0.6, 0.7]);
let distance = prosody_distance(&style_a, &style_b);
assert!(distance >= 0.0);

References

• Qian, K., et al. (2019). AutoVC: Zero-Shot Voice Style Transfer.
• Wang, Y., et al. (2018). Style Tokens for Expressive Speech Synthesis.
• Chen, M., et al. (2021). Adaspeech: Adaptive Text to Speech for Custom Voice.

PMAT Compliance

• Zero unwrap() calls
• All public APIs return Result<T, E> where fallible

Structs

AutoVcTransfer

AutoVC-based voice style transfer.

GstEncoder

Global Style Token (GST) based style encoder.

StyleConfig

Configuration for voice style transfer.

StyleVector

Voice style vector capturing prosody, timbre, and rhythm.

Traits

StyleEncoder

Trait for style encoding from audio.

StyleTransfer

Trait for voice style transfer.

Functions

average_styles

Average multiple style vectors.

prosody_distance

Compute prosody distance between two styles.

style_distance

Compute total style distance (Euclidean).

style_from_embedding

Create style from speaker embedding (approximate).

timbre_distance

Compute timbre distance between two styles.