GradTTSOptions
Options & Configuration · AiDotNet.TextToSpeech.Classic
Options for Grad-TTS (score-based diffusion decoder on a Glow-TTS-style text encoder).
Inheritance: ModelOptions → TtsModelOptions → AcousticModelOptions → GradTTSOptions
For Beginners
These options configure the GradTTS model. Default values follow the original paper settings.
How It Works
Defaults are the paper’s (Popov et al. 2021, §3–4) with unstated values from the reference implementation
(huawei-noah/Speech-Backbones, Grad-TTS/params.py): the Glow-TTS encoder (width 192, 6 relative-position
layers, 2 heads, window 4, filter 768, kernel 3, dropout 0.1; pre-net of 3 kernel-5 convolutions; duration
predictor filter 256); a U-Net score network with base width 64 and multipliers (1, 2, 4), time embedding scaled by
1000; noise schedule β_t = 0.05 + (20 − 0.05) t on t ∈ [0, 1]; 2-second training segments (172 frames);
22.05 kHz audio, 1024-point STFT, hop 256, 80 mel bins; inference with τ = 1.5 and N = 10 reverse steps.
Constructors
| Constructor | Summary |
|---|---|
GradTTSOptions() |
Creates the paper’s configuration. |
GradTTSOptions(GradTTSOptions other) |
Initializes a new instance by copying from another instance. |
Properties
| Property | Summary |
|---|---|
double BetaMax { get; set; } |
Gets or sets β₁ of the linear noise schedule (20). |
double BetaMin { get; set; } |
Gets or sets β₀ of the linear noise schedule (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 terminal-distribution temperature τ at inference (1.5). |
double TimePositionScale { get; set; } |
Gets or sets the scale of the diffusion time in its sinusoidal embedding (1000). |
int DurationPredictorFilterChannels { get; set; } |
Gets or sets the duration predictor’s filter (256). |
int EncoderDim { get; set; } |
The text encoder’s actual hidden width (192 in the released configuration). |
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 NumDecoderLayers { get; set; } |
The number of U-Net resolutions, represented by DecoderDimMultipliers. |
int NumDiffusionSteps { get; set; } |
Gets or sets the reverse-diffusion steps N at inference (10; the paper evaluates 4, 10, 100, 1000). |
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 encoder attention’s maximum relative position (4). |
int SamplingSeed { get; set; } |
Gets or sets the seed of the training draws and the inference sample, for repeatable runs. |
int SegmentFrames { get; set; } |
Gets or sets the training segment length in frames (2 seconds at 22.05 kHz / 256 = 172). |
int[] DecoderDimMultipliers { get; set; } |
Gets or sets the U-Net’s width multipliers per resolution (1, 2, 4); the base width is DecoderDim (64). |