GlowTTSOptions

Options & Configuration · AiDotNet.TextToSpeech.Classic

Options for Glow-TTS (flow-based TTS with monotonic alignment search).

Inheritance: ModelOptions → TtsModelOptions → AcousticModelOptions → GlowTTSOptions

For Beginners

These options configure the GlowTTS model. Default values follow the original paper settings.

How It Works

Defaults are the paper’s LJSpeech configuration (Kim et al. 2020, §3.3, §4, App. A.1, Table 3) as released in the reference implementation’s configs/base.json: encoder width 192, 6 relative-position Transformer layers with 2 heads, maximum relative position 4, feed-forward filter 768 with kernel 3, dropout 0.1; a pre-net of three kernel-5 convolutions with dropout 0.5; a duration predictor with filter 256; a decoder of 12 flow blocks over pairs of frames squeezed into 160 channels, each an ActNorm, an invertible 1×1 convolution mixing groups of 4 channels and an affine coupling layer with a 4-layer WaveNet-like network (width 192, kernel 5, dilation 1, dropout 0.05); 22.05 kHz audio, 1024-point STFT, hop 256, 80 mel bins; sampling temperature 0.333 (§5.1).

Constructors

Constructor Summary
GlowTTSOptions() Creates the paper’s configuration.
GlowTTSOptions(GlowTTSOptions other) Initializes a new instance by copying from another instance.

Properties

Property Summary
double DecoderDropout { get; set; } Gets or sets the coupling networks’ dropout (0.05).
double LengthScale { get; set; } Gets or sets the duration multiplier at inference (1).
double PrenetDropout { get; set; } Gets or sets the pre-net’s dropout (0.5).
double Temperature { get; set; } Gets or sets the prior’s sampling temperature at inference (0.333, the paper’s best).
int CouplingLayers { get; set; } Gets or sets the gated layers per coupling network (4).
int DecoderDilationRate { get; set; } Gets or sets the coupling layers’ dilation base (1).
int DecoderDim { get; set; } The native affine coupling network’s hidden width (192).
int DecoderHiddenChannels { get; set; } Gets or sets the width of the coupling layers’ WaveNet-like networks (192).
int DecoderKernelSize { get; set; } Gets or sets the coupling layers’ convolution kernel (5).
int DurationPredictorFilterChannels { get; set; } Gets or sets the duration predictor’s filter (256).
int EncoderDim { get; set; } The native text encoder’s hidden width (192).
int EncoderKernelSize { get; set; } Gets or sets the encoder feed-forward and duration predictor kernel (3).
int FilterChannels { get; set; } Gets or sets the encoder feed-forward filter (768).
int InvertibleConvGroupSize { get; set; } Gets or sets the channel group size of the invertible 1×1 convolutions (4).
int NumDecoderLayers { get; set; } The number of native decoder flow blocks (12).
int NumFlowBlocks { get; set; } Gets or sets the decoder’s flow blocks (12).
int PrenetKernelSize { get; set; } Gets or sets the pre-net’s kernel (5).
int PrenetLayers { get; set; } Gets or sets the pre-net’s convolutions (3).
int RelativeWindow { get; set; } Gets or sets the maximum relative position of the encoder’s attention (4).
int SamplingSeed { get; set; } Gets or sets the seed of the prior sample drawn at inference, so synthesis is repeatable.
int SqueezeFactor { get; set; } Gets or sets how many frames the decoder squeezes into channels (2).