Tacotron2Options

Options & Configuration · AiDotNet.TextToSpeech.Classic

Options for Tacotron 2 (location-sensitive attention with WaveNet vocoder).

Inheritance: ModelOptions → TtsModelOptions → AcousticModelOptions → Tacotron2Options

For Beginners

These options configure the Tacotron2 model. Default values follow the original paper settings.

Constructors

Constructor Summary
Tacotron2Options() Initializes a new instance with default values.
Tacotron2Options(Tacotron2Options other) Initializes a new instance by copying from another instance.

Properties

Property Summary
double ConvolutionDropout { get; set; } Gets or sets the dropout after every encoder and post-net convolution (0.5, §2.2).
double PrenetDropout { get; set; } Gets or sets the decoder pre-net’s dropout, applied in training and at inference (0.5, §2.2).
double StopThreshold { get; set; } Gets or sets the inference stop-token probability threshold.
double ZoneoutProbability { get; set; } Gets or sets the decoder LSTMs’ zoneout probability (0.1, §2.2).
int AttentionDimension { get; set; } Gets or sets the attention projection dimension.
int AttentionKernelSize { get; set; } Gets or sets the location convolution’s odd filter length (31, §2.2).
int AttentionLocationChannels { get; set; } Gets or sets the attention location feature channels.
int AttentionRnnDim { get; set; } Gets or sets the attention RNN dimension.
int DecoderRnnDim { get; set; } Gets or sets the decoder RNN dimension.
int EmbeddingDim { get; set; } Gets or sets the character embedding width (512, §2.2).
int PrenetDim { get; set; } Gets or sets the prenet dimension.
int SamplingSeed { get; set; } Gets or sets the seed of the inference pre-net dropout, so the same text synthesizes the same output.