VITS TTS
The end-to-end TTS architecture that combines a variational autoencoder, normalizing flows, and adversarial training.
Ojehaijey ñe'ẽnguéra etiquetas SSML-pe peteĩ control hekopete g̃uarã:
<speak><prosody rate="slow">Slow speech</prosody></speak>
Etiquetas ohechakuaáva modelo ojeporavóva - tesãirã peteĩ peteĩva ñe'ẽnguérape, oĩhápe:
Ko modelo ohai texto ndahasyivéva, upévare umi etiqueta oĩva línea ryepýpe ndojehechakuaái. Umi emoción oñemopyendáva etiqueta-pe g̃uarã, oñemoambue peteĩ modelo expresivo-pe taha'e Orfeo térã Bark.
Oñemohenda ñe'ẽnguéra ojehechapyréva (tembiapo = ñe'ẽnguéra):
Mba'épa 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.
Oñeha'ãvéva: General-purpose text-to-speech with natural prosody
Ojehecha opavave VITS ñe'ẽPeteĩ jehecha
- Desarrollador
- Jaehyeon Kim et al.
- Licencia
- MIT
- Ta'ãnga
- free
- Velocidad
- fast
- Clonación ñe'ẽnguéra rehe
- No
- Ñe'ẽ
- English, German, Spanish, French, Portuguese, Dutch, Finnish, Hungarian, Bulgarian, Japanese, Polish
- Caracteres máx.
- 2000