search

TensorSpeech / TensorFlowTTS

:stuck_out_tongue_closed_eyes: TensorFlowTTS: Real-Time State-of-the-art Speech Synthesis for Tensorflow 2 (supported including English, French, Korean, Chinese, German and Easy to adapt for other languages)
https://tensorspeech.github.io/TensorFlowTTS/
Apache License 2.0
3.83k stars 813 forks source link

My Esperanto TTS is overfitting #694

Closed riproskaie closed 2 years ago

riproskaie commented 3 years ago

Hello, I've been working on implementing an Esperanto fork of TensorflowTTS, but for the past few days, I haven't been so successful. My 6k-trained model spits random scrapped sounds (about 46 seconds long) from my audio input. I gave the model a bigger dataset, and it now talks nonsense for about 5~10 seconds for my 15-character input string. This is much shorter, but I'm not sure I can call it an improvement.

My current Esperanto dataset is 50-minute long. I made sure the cleaner is processing my strings right, and there were no typos in my metadata. Should I keep training the model for additional n-k steps, or do I need more audio recordings?

Here is my Tensorboard:

image

dathudeptrai commented 3 years ago

@ZDisket can you help him, seems 50 - minute dataset is not enough for tacotron ?

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.