diff --git a/docs/source/tts/demo.rst b/docs/source/tts/demo.rst index ca2fd98e4..c7aa4722d 100644 --- a/docs/source/tts/demo.rst +++ b/docs/source/tts/demo.rst @@ -1735,4 +1735,142 @@ We use ``FastSpeech2`` + ``ParallelWaveGAN`` here.

- \ No newline at end of file + +Finetune FastSpeech2 for CSMSC +-------------------------------------- + +Finetuning demos of https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/other/tts_finetune/tts3 for CSMSC dataset. + +When finetuning for CSMSC, we thought ``Freeze encoder`` > ``Non Frozen`` > ``Freeze encoder && duration_predictor`` for audio quality. + +.. raw:: html + +
+ CSMSC reference audio(fastspeech2_csmsc + hifigan_aishlle3 in CLI): 欢迎使用飞桨语音套件。 +
+
+ +
+
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Frozen Method num=10, bs=10,
epoch=100, lr=1e-4 		num=18, bs=18,
epoch=100, lr=1e-4 		num=97, bs=64,
epoch=100, lr=1e-4 		num=196, bs=64,
epoch=100, lr=1e-4
Non Frozen + + + + + + + +
Freeze encoder + + + + + + + +
Freeze encoder &&
duration_predictor		 + + + + + + + +
+ +
+