NaturalSpeech2Diffusion

Enums · AiDotNet.TextToSpeech.CodecBased

What NaturalSpeech 2’s latent diffusion learns.

Fields

Field Summary
NaturalSpeech2Diffusion FlowMatching Amphion’s alternative: rectified flow, z_t = (1 − t) z₀ + t ε, the denoiser predicting ε − z₀.
NaturalSpeech2Diffusion ScoreSde The paper’s variance-preserving SDE with a data-predicting denoiser (Eqs.
int value__