LjSpeechDataLoaderOptions
Options & Configuration · AiDotNet.Data.Audio.Benchmarks
Configuration options for the LJSpeech 1.1 data loader (Ito & Johnson 2017).
How It Works
LJSpeech is the canonical single-speaker English TTS corpus — 13,100 audiobook clips (≈ 24 hours) from a single female narrator at 22,050 Hz. Used as the default TTS training corpus by Tacotron, FastSpeech, VITS, and most subsequent neural TTS papers.
Constructors
| Constructor | Summary |
|---|---|
LjSpeechDataLoaderOptions() |
Properties
| Property | Summary |
|---|---|
DatasetSplit Split { get; set; } |
Dataset split. |
bool AutoDownload { get; set; } |
Gets or sets whether the dataset is downloaded into the data directory when it is not there. |
bool UseNormalizedText { get; set; } |
Use the normalized text (numbers/dates expanded). |
int MaxAudioSamples { get; set; } |
Maximum waveform length in samples (zero-padded). |
int MaxTextLength { get; set; } |
Maximum transcript length in tokens (input text). |
int VocabularySize { get; set; } |
Vocabulary size for the text tokenizer. |
int? MaxSamples { get; set; } |
Gets or sets the most samples loaded; null loads the whole split. |
string DataPath { get; set; } |
Gets or sets the dataset root directory; null uses the default data directory for this dataset. |
Methods
| Method | Summary |
|---|---|
void Validate() |
Validates that all option values are within acceptable ranges. |