deepvoice3

deepvoice3 copied to clipboard

Hello,

You are amazing. training really super fast than tactron. but after successfully trained 256k model i am only getting noise in output wav file generated using synthesize.py.

I am not expert like you on speech synthesizing can you please help me out for this? i am using aws deep learning with GPU machine

Here is the attached wav and graph file let me know if you want anything else.

Any suggestion would be highly appreciated :) .

data.zip

Thanks

The best suggestion I can do is to consider other models like tacotron or dctts. You can find their implementation in my github. They are much easier to train.

Hello Kyubyong,

Thanks for the reply. means my trained model are useless? right or i can train more using deepvoice3 ?

Also i would like to know if i use your other implementation like tacotron or dctts. then how many steps of training it will required to get satisfied result?

My first choice is your deepvoice3 project and if you can guide me to fix the issue that will be great.

Could you at least provide why its only generating noise? im experiencing similar issues. at the time of writing this im still training until i meet at least the 500k epochs as defined in the hparams. my hypothesis here is that there's just not enough training done. am i on the right path or am i wasting my time, training? ^^

I'm at 600k iterations now. still just noise. my current hypothesis is that there is something wrong with the preprocessor or the synthesizer, which i cant solve without intimate knowledge of how this algorithm works. and since im a mere beginner in the field of data science, im giving up on this project. it seems to be dead anyway.