PaddlePaddle / PaddleSpeech

Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation and Keyword Spotting. Won NAACL2022 Best Demo Award.
https://paddlespeech.readthedocs.io
Apache License 2.0
10.99k stars 1.83k forks source link

[TTS]example/aishell3/tts3 自己的数据集训练synthesize_e2e出来没有声音 #2574

Closed HandsLing closed 1 year ago

HandsLing commented 1 year ago

请问在example/aishell3/tts3 中,我用aishell3数据集训练没有问题,但是我用我自己的数据训练,不报错,但是推理出来的音频没有声音,只有 “”enenen“”的声音,请问这是怎么回事呢

yt605155624 commented 1 year ago

请检查

  1. synthesize 时是否可以正常合成吗,如果可以证明训练的模型没有问题
  2. synthesize_e2e 时候输入的 phone_id_map.txt 和训练时是否一致
HandsLing commented 1 year ago

不好意思,是我这边没有把phone_id_map.txt这个路径给对导致的问题,换了正确的路劲就没问题了,谢谢