Testing Voice Model (CPU)
Last updated
Was this helpful?
Last updated
Was this helpful?
Alright now that you've done training your voice model it's time to test your voice model
Be sure to copy to your drive.
Alright so in this main page, you might be confused on where to's. Don't worry, we'll follow everything like it says in a colab page. So let's go for step by step.
This will setup everything it needs (e.g downloading resources needs like tacotron2 and HIFI-GAN, etc)
Now this is where the main part starts. First off, you need get a voice model that you trained, go to your google drive and select this folder:
And after that you right click the model that you made:
After that, copy that ID, NOT THE ENTIRE LINK! THE ONE WITH AN ID ON IT THAT STARTS WITH "https://drive.google.com/file/d/*YOUR ID*/view?usp=sharing"
And then you paste it in tacotron_id: *YOUR ID*
Now in second section you won't need it, it's really that optional, unless if you want to make it sound better, then only if you have fine-tuned model to it, which we'll do that later on.
If you did fine-tuned model, do the same thing with google drive's id link to it.
Otherwise, keep universal.
The third option is pronounciation_dictionary.
If you have ARPABETS in your voice model then better check that, otherwise keep it unchecked if your transcript is english and not ARPABETS.
Fourth option is show_graphs.
Really optional if you wanna see some results yourself (i recommend keeping that on because you may need some of that graph.)
Finally, max_duration.
Depends how long you want your wav file to be, but don't put it too long or else it might synthesis long, or it'll just not play it for you and show an error.
Now once that's all with it click play and wait. Until you'll see this:
Just type something or maybe grab some copypasta until you'll see this:
If you happy with the results then congratulations! You can now upload to uberduck now!
If your model sounds bad nor even spectrogram is just wonky then keep training or grab some more voice files until it gets better.
Congratulations, you have successfully made an voice model. Again you can upload it to uberduck or just keep it private on your own.
If you have some errors go ahead for the next page of Troubleshooting or always ask in Uberduck's discord server.
