Google · 2017 · seq2seq

Tacotron

Characters in, spectrograms out, trained from scratch. Tacotron is the paper that made end-to-end neural TTS a research default. The clip is the authors’ out-of-domain demo; Tacotron 2 reads the same sentence.

Known for

Ending the era of hand-built TTS frontends.

Vendor / paper demo

Tacotron paper demo · Tacotron official demo, gan_or_vae_r2.wav. Same sentence as the Tacotron 2 clip.

Generative adversarial network or variational auto-encoder.

Approach
AR seq2seq
Year / era
2017
License class
research
Cloning
No
Streaming
No
Emotion controls
No
Languages (tags)
en
Official
Source
Kokoro 82M · af_heart