DiTToTTSOptions

Options & Configuration · AiDotNet.TextToSpeech.FlowDiffusion

Options for DiTTo-TTS. Defaults are the English Base system from Lee et al. (2024): frozen microsoft/speecht5_tts text features, CLaM-TTS’s 512-wide 10.76 Hz Mel-VAE latent, the 12-layer/768-wide DiT, the separate length predictor, and 25-step classifier-free DDIM.

Inheritance: ModelOptions → TtsModelOptions → EndToEndTtsOptions → DiTToTTSOptions

Constructors

Constructor Summary
DiTToTTSOptions() Creates an instance with the default settings.
DiTToTTSOptions(DiTToTTSOptions other) Copies the settings from another options instance.

Properties

Property Summary
DiTToCodecTrainingOptions CodecTraining { get; set; } Gets or sets the codec training settings.
DiTToTrainingStage TrainingStage { get; set; } Gets or sets the stage to optimize during training.
double ClassifierFreeDropProbability { get; set; } Gets or sets the classifier free drop probability.
double FullMaskProbability { get; set; } Gets or sets the full mask probability.
double GuidanceScale { get; set; } Gets or sets the guidance scale.
double MaximumMaskRatio { get; set; } Gets or sets the maximum mask ratio.
double MinimumMaskRatio { get; set; } Gets or sets the minimum mask ratio.
double NoiseScheduleScale { get; set; } Gets or sets the noise schedule scale.
int BigVganInitialChannels { get; set; } Gets or sets the BigVGAN initial channels.
int DiTFeedForwardMultiplier { get; set; } Gets or sets the diffusion transformer feed forward multiplier.
int DiTHeads { get; set; } Gets or sets the number of attention heads in the diffusion transformer.
int DiTHiddenDim { get; set; } Gets or sets the feature width of the diffusion transformer hidden.
int DiTLayers { get; set; } Gets or sets the number of layers in the diffusion transformer.
int FlowDim { get; set; } Compatibility name used by the old placeholder.
int InferenceSteps { get; set; } Gets or sets the inference steps.
int LatentDim { get; set; } Gets or sets the feature width of the latent.
int LengthFeedForwardDim { get; set; } Gets or sets the feature width of the length feed forward.
int LengthHeads { get; set; } Gets or sets the number of attention heads in the length.
int LengthHiddenDim { get; set; } Gets or sets the feature width of the length hidden.
int LengthLayers { get; set; } Gets or sets the number of layers in the length.
int LengthTopK { get; set; } Gets or sets the length top k.
int MaxLatentLength { get; set; } The maximum remaining-length class N.
int MelVaeChannels { get; set; } Gets or sets the mel VAE channels.
int MelVaeDownsampleFactor { get; set; } Gets or sets the mel VAE downsample factor.
int NumFlowLayers { get; set; } Compatibility name used by the old placeholder.
int RvqDepth { get; set; } Gets or sets the residual vector quantizer depth.
int RvqVocabulary { get; set; } Gets or sets the residual vector quantizer vocabulary.
int SamplingSeed { get; set; } Gets or sets the random seed used during sampling.
int TextEncoderLayers { get; set; } Gets or sets the number of layers in the text encoder.
int TextEosId { get; set; } Gets or sets the token ID for text eos.
int TextFeedForwardDim { get; set; } Gets or sets the feature width of the text feed forward.
int TextHeads { get; set; } Gets or sets the number of attention heads in the text.
int TextHiddenDim { get; set; } Gets or sets the feature width of the text hidden.
int TextPadId { get; set; } Gets or sets the token ID for text pad.
int TextVocabulary { get; set; } Gets or sets the text vocabulary size.
int TrainingDiffusionSteps { get; set; } Gets or sets the training diffusion steps.
int[] BigVganResidualDilations { get; set; } Gets or sets the convolution dilations for the BigVGAN residual.
int[] BigVganResidualKernels { get; set; } Gets or sets the convolution kernel sizes for the BigVGAN residual.
int[] BigVganUpsampleKernels { get; set; } Gets or sets the convolution kernel sizes for the BigVGAN upsample.
int[] BigVganUpsampleRates { get; set; } Gets or sets the BigVGAN upsample rates.
int[] MelVaeMultipliers { get; set; } Gets or sets the mel VAE multipliers.
string SentencePieceModelPath { get; set; } Gets or sets the path to the SentencePiece tokenizer model.

Methods

Method Summary
DiTToTTSOptions UseSize(DiTToModelSize size) Applies one of the exact DiT size rows in Table 1.