Uberduck

Nice. if it gets accepted when i submit it ill work on the next model.

Bobs burgers is my dad‘s favourite programme

oh great `download_from_google_drive()` in the tacotron notebook just broke somehow

```
IndexError                                Traceback (most recent call last)

<ipython-input-4-8282cd122340> in <module>()
      1 #@markdown #/mlp/ dataset loader
      2 google_drive_id = "1R-A7FO1zHGLUGbsvU8JWFG8kcJ44d1QE" #@param {type:"string"}
----> 3 download_from_google_drive(google_drive_id, "dataset.tar")
      4 os.mkdir("/content/tacotron2/wavs2")
      5 get_ipython().system('tar -xvf dataset.tar -C /content/tacotron2/wavs2')

<ipython-input-3-0682750d4349> in download_from_google_drive(file_id, file_name)
     57   get_ipython().system('curl -c ./cookie -s -L "https://drive.google.com/uc?export=download&id={file_id}" > /dev/null')
     58   confirm_text = get_ipython().getoutput("awk '/download/ {print $NF}' ./cookie")
---> 59   confirm_text = confirm_text[0]
     60   get_ipython().system('curl -Lb ./cookie "https://drive.google.com/uc?export=download&confirm={confirm_text}&id={file_id}" -o {file_name}')
     61 

IndexError: list index out of range
```

loading the sweetie belle dataset (`13RdwFLT_-iiXjLL9MS6LvUVlBv90QBig`) works but sunset shimmer (`1R-A7FO1zHGLUGbsvU8JWFG8kcJ44d1QE`) doesn't
note that i did sweetie belle earlier but in a separate session

i factory reset the colab a few times already

after further checking it seems that it legit only lets me download files i've already downloaded in previous sessions

I have no idea how to code. Where’s a good place to start learning? (I want to stuff like ai and game dev) :)

is this the danger of voice cloning that people have been warning us about

https://tenor.com/view/tyler1-dead-ko-faint-gif-17440398

hold up how do I make a finetuned hifi-gan model

I think that could make old coach (rec room) sound pretty good

as old coach sounds good even with 50 audio files

probably safe to say the experiment didn't go as well as i thought

hmm how did this happen (the new gianni duke model)
it sounds like spleeter artifacts even though the dataset doesn't use spleeter

the bottommost notebook in <#841437191073955920>
and then for testing you have to manually add it into the code of the hifi-gan synthesis notebook
though idk if uberduck even supports finetuned hifi-gan currently

i tried making billie eilish say "baby i'm drunk and i don't wanna go home" but it just sounds like she is crying lol

The Billie Eilish voice is pretty broken on Uberduck

hm apparently i have a 2.66gb file on my google drive called `checkpoint_run1.tar`
what's weird is it's from before i even heard about uberduck

only thing i can think of that stores checkpoints in .tar archievs is gpt2