Crate voirs_vocoder 

VoiRS Vocoders

Neural vocoders for converting mel spectrograms to high-quality audio. Supports HiFi-GAN, WaveGlow, and other state-of-the-art vocoders.

This crate provides a unified interface for neural vocoding with support for:

• Multiple neural architectures (HiFi-GAN, DiffWave, WaveGlow)
• Real-time streaming synthesis
• Batch processing for efficiency
• Quality control and performance monitoring
• Singing voice processing with harmonic enhancement
• Spatial audio processing with HRTF

Quick Start

// This is a conceptual example - actual usage will depend on specific implementations
use voirs_vocoder::{Vocoder, MelSpectrogram, SynthesisConfig};

async fn example_usage() {
    // Create a mel spectrogram (this would typically come from a TTS model)
    let mel_data = vec![vec![0.0; 80]; 100]; // 80 mel bands, 100 time steps
    let mel = MelSpectrogram::new(mel_data, 22050, 256);
     
    // Configure synthesis parameters
    let config = SynthesisConfig::default();
     
    // Convert to audio using any Vocoder implementation
    // let audio_buffer = vocoder.vocode(&mel, Some(&config)).await?;
}

Re-exports

pub use hifigan::HiFiGanVocoder;

pub use models::hifigan::HiFiGanConfig;

pub use models::hifigan::HiFiGanVariant;

pub use models::hifigan::HiFiGanVariants;

pub use streaming::StreamHandle;

pub use streaming::StreamingPipeline;

pub use streaming::StreamingStats;

pub use streaming::StreamingVocoder;

Modules

adaptive_quality

Adaptive quality control system for vocoders

audio

Audio processing module for voirs-vocoder.

backends

Backend infrastructure for voirs-vocoder.

broadcast_quality

Professional broadcast quality enhancement for VoiRS

cache

Cache optimization module for improved data locality and memory access patterns

codecs

Audio codec implementations.

comprehensive_quality_metrics

Comprehensive quality metrics for all vocoder features

conditioning

Unified conditioning interface for advanced vocoder features.

config

Configuration system for voirs-vocoder.

containers

Audio container format implementations.

conversion

Voice conversion functionality for real-time voice transformation.

drivers

Real-time audio drivers for voirs-vocoder

effects

Audio effects and enhancement system for voirs-vocoder.

hifigan

Auto-generated module structure

loss

Perceptual loss functions for neural vocoder training and evaluation.

metrics

Audio quality metrics for vocoder evaluation

ml

Machine Learning Enhancement Module

models

Vocoder model definitions and implementations.

optimization_paths

Feature-specific optimization paths for vocoder operations

parallel

Parallel processing utilities for voirs-vocoder

performance

Real-time performance monitoring and quality metrics tracking

post_processing

Post-processing pipeline for audio enhancement and quality improvement

prelude

Prelude for convenient imports

profiling

Performance profiling utilities for vocoder operations

simd

SIMD acceleration for audio processing operations

streaming

Real-time streaming architecture for neural vocoders

utils

Utility modules for vocoder operations

waveglow

WaveGlow vocoder implementation.

Structs

AudioBuffer

Audio buffer for holding PCM audio data

DummyVocoder

Dummy vocoder for testing

MelSpectrogram

Mel spectrogram representation

SynthesisConfig

Synthesis configuration

VocoderManager

Vocoder manager with multiple architecture support

VocoderMetadata

Vocoder metadata

Enums

LanguageCode

Language codes supported by VoiRS

VocoderError

Vocoder-specific error types

VocoderFeature

Features supported by vocoders

Traits

Vocoder

Trait for neural vocoders

Type Aliases

Result

Result type for vocoder operations