[Feature request]

[kunge98]

My code is shown in the figure, using the model of Chinese speech synthesis, my synthesis test text is "你好", when I open the output audio, I find that the audio has 4 seconds, but the two words "你好" obviously do not need that long time, and the synthesized audio is followed by some sounds similar to howling, may I ask what I need to do? Do you need to modify the config.json file?

[kunge98]

Then I tried to modify the max_decoder_steps value in the config.json file, and the output voice confirmation could be shortened. But right now I want to do speech generation for bulk text, but the length of this article is different. How can I do that?