hmm

not a

.RAR

i will fix that right now and put you on the bugfinder board

there

refresh and try again

thanks 🙂 I'm sorry that because of a noob like me, you have to waste your time on this xD

Hooray! I managed to finish the second stage. Now I'll start to move on to the next ones, and I hope I succeed 😅

Now I have encountered a strange thing. I finished stage 4, and to the left of the stage activation I have a green checkmark, but at the end of the list I see a strange inscription with an error, and does this mean that something went wrong? Or I shouldn't worry since there is a green checkmark to the left of the activation of the stage

it should keep running so yeah that's wrong

send

it's not like i'd spend it on anything better tbh

and at the end this "RuntimeError: shape '[1, 1, 252345]' is invalid for input of size 504690"

then I can worry a little less xD

afaik that means your audio is either 44khz or stereo

go back to step 2, turn on

autofix

, and load again

aahhh

Should I edit each of my audio files in Audacity? Convert to 22050?

Yes

Then I understood, thank you

Now I see this process. This file in the arrow is what I was able to achieve by traning from my files?

I turn it on and the 2 files above and they sound especially raucous compared to their originals. Is it because their frequency is 44100 HZ? And at 22050 the sound of the files will be without hoarseness?

no the hertz were already taken care of

it sounds warbly because the vocoder used here is a very simple one

aahh, hmm

you can use the synthesis notebook when you're done to check the quality

hmm ok, I'll try

for the tacotron legacy notebook, how do you make it re-compatible with the current version of google collab?

Is my version of the notebook incompatible?

i'll look again, last time i ran it, some package was erroring out because it was depreciated

I'm training a 6 minutes dataset so what should i set the epoch, batch size, and the learning rate as.