VITS TTS
The end-to-end TTS architecture that combines a variational autoencoder, normalizing flows, and adversarial training.
Wrap din tekst i SSML tags for præcis kontrol:
<speak><prosody rate="slow">Slow speech</prosody></speak>
Tags den valgte model forstår! klik for at droppe en i din tekst, hvor det sker:
Denne model læser almindelig tekst, så inline tags ignoreres. For tag-baserede følelser, skifte til en ekspressiv model som Orpheus eller Bark.
Definer brugerdefinerede udtaler (ord = udtale):
Om VITS
VITS — Variational Inference with adversarial learning for end-to-end Text-to-Speech — was introduced by Jaehyeon Kim and collaborators in 2021 and became a foundational architecture for modern neural speech. Rather than the older two-stage pipeline, it synthesizes audio in a single parallel end-to-end pass, pairing a variational autoencoder with normalizing flows and a GAN-style adversarial training process to lift naturalness. At about 25M parameters and trained on ~585 hours, it produces natural prosody at fast inference speeds and supports multiple speakers. It serves as a solid general-purpose, free baseline and underpins many later models such as Piper and MeloTTS.
Bedst for: General-purpose text-to-speech with natural prosody
Gennemse alle VITS stemmerEt blik
- Udvikler
- Jaehyeon Kim et al.
- Licens
- MIT
- Metodetrin
- free
- Hastighed
- fast
- Stemmekloning
- Nej
- Sprog
- English, German, Spanish, French, Portuguese, Dutch, Finnish, Hungarian, Bulgarian, Japanese, Polish
- Maks. tegn
- 2000