使用CSMSC数据集训练了transformer声源模型，waveflow作为声码器进行语音合成

[1726650686]

PR types

PR changes

Describe

使用CSMSC数据集训练了transformer声源模型，waveflow作为声码器进行语音合成

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

Lv Zhixuan seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[yt605155624]

感谢贡献，音频特征提取 （fs/n_fft/win_length/n_shift/fmin/fmax/n_mels/window）参数可以与https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/csmsc/tts3/conf/default.yaml 保持一致，这样就可以使用 csmsc 的声码器了，目前 waveflow 只有在 ljspeech 上的模型，并且不打算维护（速度慢，不流行）

[yt605155624]

麻烦签署一下 CLA 吧

[yt605155624]

可以扫描主页二维码加用户群，艾特我（子龙）详聊

[yt605155624]

提交代码按照 https://github.com/PaddlePaddle/PaddleSpeech/issues/2325 用 pre-commit 刷一下代码格式，否则过不了 CodeStyle 这一条 CI

[mergify[bot]]

This pull request is now in conflict :(