Back: Catalog

VITS

The project is a method of synthesizing from a conditional variational autocoder-based end-to-end text (TTS) using adversarial learning. It produces more natural sounds through a method of variational inference supported by normalizing flows and the contentious training process. It contains a stochastic time estimator to synthesize speech at different rhythms than the introductory text.

VI

VITS

48

jaywalnut310 · Voice & Text-to-Speech

STARS7.9K
FORKS1.4K
COMMITS12
CONTRIBUTORS3
VERSION—
UPDATED2 yr ago
PythonConditional VAEEnd-To-EndResearchMIT

Common commercial products in this category: ElevenLabs, Amazon Polly, Google Cloud Text-to-Speech

What can it do?
  • You perform model training using LJ Speech and VCTK datasets
  • Using the pre-trained models provided and the Colab notebook, you synthesize interactive sound
  • Convert text data to audio files via the Inference instance
Heads up
  • • Last update 2 years ago
User reviews
  • ↓ NegativeGitHub Issue

    Result getting worse when i use ground truth duration.

    “Dear author, thank you for your contribution for TTS, this is a big step in E2E TTS. But when I use ground truth duration aiming to train faster and get more accurate duration, duration loss drops fast and kl loss drops slow. I only change the attn matrix using true duration. I…”

  • • NeutralGitHub Issue

    is that able to train on Chinese dataset?

    “is that able to train on Chinese dataset?”

Website GitHub

My note

Alternative pages

ElevenLabs alternativesAmazon Polly alternativesGoogle Cloud Text-to-Speech alternatives

More in: Voice & Text-to-Speech