Testing Voice Model

This is synthesis page for GPU version, but if you want to use CPU version please go here: CPU Version

Colab page

Be sure to copy to drive.

Usage synthesis page

Read all this stuff if you want.

Scroll down till we'll find this section:

What are these?

Okay, first things first, let's start of one by one section:

• character - Basically just a list of characters that have fine-tuned with HIFI Gan.

• tacotron_override - This is the one that you should be doing on your own model, which we will be covering in this guide

• pronounciation_dictionary - Your model might go gibberish, so I recommend do uncheck that. It reads through arpabets of that.

• show_graphs - Keep checked. It shows your mel and graphs.

• max_duration - Determines how many seconds it'll be played audio. I recommend pulling 10 but you can put more, but not recommended.

• stop_threshold - I recommend not to touch that since I never touched this before

• superres_strength - I also recommend not to touch that.

So now that we've covered everything let's get things started with tacotron_override.

Loading your model

So what you want is to go to your google drive and then do this. colab --> outdir folder:

You should have these models that doesn't go .zip, if you haven't trained your models then what are you doing? go back to the training section!

Training time

Once you are in there, go ahead and right click model and select "Get link":

You should be then greeted by this pop-up:

Yours might be different but select restricted into anyone with the link.

Once you have done, here's what you should copy for:

https://drive.google.com/file/d/*google drive id*/view?usp=sharing

Google drive id, is the one you should be copying for. Go ahead copy then go back to colab page and paste it on tacotron_override:

Now once that's done, run this cell and wait until it downloads some parts.

Synthesis fun time

Now once it has done things, go ahead and type everything you need and you should see this:

If everything has done correctly, then congratulations. Your model is now ready for Uberduck!

Sharing your model into Uberduck

Now that you've tested your model, you can now go ahead share your model with the public via website or discord server.

Then you gonna have to wait until mods approve your model.

If your model gets approved you should see it on #announcements so that if your model got added to Uberduck.

If not, then wait for it. Patience is the key.

Finish

Congratulations, you have now made your own voice model for your favorite character or your own voice. Enjoy your trained model and be sure to share it with Uberduck if you want your model to be public.