[Feature]: 【TTS】The source audio has no interference, the generated audio is blurry, and the pronunciation of individual characters is correct

[DecodeW]

Is your feature request related to a problem? Please describe.

Creating autio by using tts，some can be correct，others are not blurry，even isn't human language. 10 words only 2-3 correct.

Describe the solution you'd like

1.generate a normal audio and no strange voice.

Describe alternatives you've considered

1.I edited the dot format to match language and blank,it has a little improvement.But i think it is not the main problem.
2.I changed target_len to different value and None, no useful.

Additional context

[yuantuo666]

Hi, MaskGCT is trained on Emilia, which means its training data is in 3-30 seconds. It seems your prompt text is a little bit lengthy and missing punctuation. Maybe you can try:

1. Cut the prompt_wav into 3-10 seconds. (Make sure the total length of prompt wav and generated wav is within 30 seconds)
2. Adding punctuation for prompt_text.

[yuantuo666]

Related issue: #305

[DecodeW]

I have tried these method that you told. The result is also same as before.Just a little improvment.