And how it's structured

I initially was training hifigan with radtts so mine looked like this:

What's the point of there being different people

Why though. To just train more stuff at a time in one runtime?

I see no point unless I'm misunderstanding it

a multispeaker dataset can speak multiple voices, but also it can generalize a bit between the speakers iirc. So like you could have a different person with singing samples help to make speaking models able to sing too

with tacotron 2 I've worked with a script that lets you blend those voices in different strengths, curious if talknet has similar

so there can be individual model performance gains (potentially) as well as interesting inter-model capabilities

how it stacks compared to a single speaker with 8 hours of well transcribed training I can't say though

I've got my model where it has singing and talking data being thrown at it at once

Would that be bad or are there any advantages to having a multispeaker system?

for an individual voice idk, trying to figure that out with this testing

when you're more constrained, like no singing for a voice exists, it definitely can provide benefits

Oh well that's good

Is there a way to make the notebooks automatically go to the next step in colab?

I don't like sitting here waiting for it to finish steps

It takes FOREVER

Especially at step 6

press shift enter to start a cell and focus on next cell, then repeat as many times as desired

i got a problem when setup TalkNet

As far as I can tell you simply need to go to runtime at the top row around "file" and then click restart runtime. Or you can just press Control+M

I managed to completely make a model, but ran out of GPU points in collab right before I tried to export it!

What can I do!

Wait I'll transfer everything to another google account with enough drive space and simply load the checkpoints. Then I can export.

hey, im using controllable talknet on google colab but i dont really understand where I know the ID of my custom model, does anyone know?

Very simple. You find where your model actaully is in your google drive (it'll be a zip) and right click it. After doing that click "share" Make the settings so that it's visibility is anybody with the link then click "get link" or "copy" (I think it's the first) then copy and paste it into the "Drive ID for custom model" in the notebook

For example, https://drive.google.com/file/d/10JcXeMNR3L4SJvlDrVNsW5tdwMV2E_of/view?usp=share_link

You take the clip of that and make it only the numbers and letters, whic would be

10JcXeMNR3L4SJvlDrVNsW5tdwMV2E_of

You put that in your field in the notebook