IEndToEndTts<T>

Interfaces · AiDotNet.TextToSpeech.Interfaces

Interface for end-to-end TTS models that generate waveforms directly from text without a separate vocoder stage.

Implements: ITtsModel<T>

Known implementations: CoMoSpeech<T>, DiTToTTS<T>, E3TTS<T>, EmotiVoice<T>, Kokoro<T>, MatchaTTS<T>, MegaTTS2<T>, MegaTTS3<T>, MegaTTS<T>, MeloTTS<T>, MetaVoice1B<T>, NaturalSpeech2<T>, NaturalSpeech3<T>, NaturalSpeech<T>, OpenVoiceV2<T>, OpenVoice<T>, StyleTTS2<T>, StyleTTSZS<T>, VitsTtsModelBase<T>, VoiceFlow<T>

How It Works

End-to-end models combine acoustic modeling and waveform generation in a single model: Text -> [Single Model] -> Waveform. Architectures include:

  • VITS: VAE + normalizing flow + adversarial training
  • VITS2: improved alignment and duration prediction
  • YourTTS: multilingual zero-shot VITS variant
  • Piper: lightweight VITS for edge deployment

Properties

Property Summary
int HiddenDim { get; } Gets the hidden dimension of the model’s internal representation.
int NumFlowSteps { get; } Gets the number of flow steps used in the posterior encoder (for VAE-based models).