Any time, just send me the data I don't want to do it on vc so i'll just record myself with obs

Sounds perfect ^^

Don't you mean what you sound like?

yes, sorry

H E L P

That's an old notebook

Use this one https://colab.research.google.com/drive/1CIPTTj94EocZe2w5-zCDC3G44K9OaFal

okay

thanks

????

You're supposed to put a zip file of list.txt + wavs folder in UPLOAD_HERE i think

Oh yeah ur right

I got an error in the Check data step and I'm not sure what I did wrong.

py
Checking Training Files
 
[WARNING] does not exist.
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-11-5fd8b979d3f3> in <module>
----> 1 check_dataset(hparams)

1 frames
<ipython-input-6-520201c45b53> in check_arr(filelist_arr)
    347             if (not os.path.exists(file[0])):
    348                 print("|".join(file), "\n[WARNING] does not exist.")
--> 349             if len(file[1]) < 3:
    350                 print("|".join(file), "\n[info] has no/very little text.")
    351             if not ((file[1].strip())[-1] in r"!?,.;:"):

IndexError: list index out of range

I think you have a blank line in your transcription

What do you mean by transcription?

I only have a txt file and wav files

You mean blank lines in the txt file? Kinda like this?

Check if you accidentally hit enter for a new line, but have nothin on that line _ _

Yea, there's all blank lines at the end in that txt file

Remove the blank line

Oh, also, I was wondering if commas and question marks are allowed and do anything.

Those will cause errors. Also minor spelling errors such as a , instead of . For "wavs/#.wav|"

They are allowed and they change the intonation to be more natural

Ah nice

Is Tacotron2 still recommended btw? Its GitHub hasn't had any updates in a long time. https://github.com/NVIDIA/tacotron2

It's old, but it's what this site uses (as well as talknet) But talknet is also old

If you want, you can use an updated version of tacotron called pipeline. The datasets are the same. Notebooks for that are in #841437191073955920

Thanks

I'm new to this and started from this tutorial.

But I've also just noticed that in this Discord channel's pinned message there's a link to a different colab for tacotron2.