VITS: variational-flow and adversarial end-to-end speech synthesis

What changed

Combined conditional variational inference, normalizing flows and adversarial waveform training with a stochastic duration predictor for parallel text-to-speech.

Scientific scope

Combined conditional variational inference, normalizing flows and adversarial waveform training with a stochastic duration predictor for parallel text-to-speech. Inherits VAEs, flows, GANs and duration/alignment methods. Delta is the integrated latent/duration model, not first end-to-end TTS or proof of full natural prosody. Ground-truth-like MOS is limited to original tested datasets/listeners.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits VAEs, flows, GANs and duration/alignment methods. Delta is the integrated latent/duration model, not first end-to-end TTS or proof of full natural prosody. Ground-truth-like MOS is limited to original tested datasets/listeners.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction