search

TensorSpeech / TensorFlowTTS

:stuck_out_tongue_closed_eyes: TensorFlowTTS: Real-Time State-of-the-art Speech Synthesis for Tensorflow 2 (supported including English, French, Korean, Chinese, German and Easy to adapt for other languages)
https://tensorspeech.github.io/TensorFlowTTS/
Apache License 2.0
3.8k stars 810 forks source link

I found that the model trained from ljspeech has no rising intonation #740

Closed ttsking closed 2 years ago

ttsking commented 2 years ago

The model trained by tacotron has no rising intonation for general question sentences like: "May I transfer you to their line?"

ttsking commented 2 years ago

And i also tested the commercial TTS like Azure TTS. Both "May I transfer you to their line?" and "May I transfer you to their line." have rising intonation at the end of sentences. I wonder that the question intonation is not determined by question mark for the TTS Model?

ttsking commented 2 years ago

I tested Azure TTS more. "And then we can talk?" has question intonation and "And then we can talk." not. So Azure TTS works fine for question intonation. What's the main point to train a Model with question intonation?

ttsking commented 2 years ago

I sovled this problem by switch model to fastspeech. But for the same training data, the tacotron cannot learn the question intonation.