https://uberduck.ai/ logo
Join Discord
Powered by
# tacotron-2-support
  • I don't think they want peeps outside the company to show the admin console.
    m
    I don't think they want peeps outside the company to show the admin console.
  • it’s an endless frustration to the point where i just gave up even attempting to approve models a long time ago
    u
    it’s an endless frustration to the point where i just gave up even attempting to approve models a long time ago
  • Plus, zwf is mainly more focused on the commercial side of things than the community one.
    m
    Plus, zwf is mainly more focused on the commercial side of things than the community one.
  • It is the one the keeps the server running.
    m
    It is the one the keeps the server running.
  • t

    https://cdn.discordapp.com/attachments/994486394049282058/1091065165472075807/image0.jpg▾

  • I don’t get this
    t
    I don’t get this
  • I’ve trained it 130 epochs now with the thing showing this
    t
    I’ve trained it 130 epochs now with the thing showing this
  • https://media.discordapp.net/attachments/994486394049282058/1091021287855497329/image0.jpg
    t

    https://media.discordapp.net/attachments/994486394049282058/1091021287855497329/image0.jpg▾

  • And it still pops up as that on the synthesizer
    t
    And it still pops up as that on the synthesizer
  • And says max decoder steps
    t
    And says max decoder steps
  • How can I fix this
    t
    How can I fix this
  • The loss is at 0.015734
    t
    The loss is at 0.015734
  • Like he says half of the line at the beginning then starts mumbling BS
    t
    Like he says half of the line at the beginning then starts mumbling BS
  • 0.01 is too high
    u
    0.01 is too high
  • go for like 0.009
    u
    go for like 0.009
  • also access denied, even though I set the link to "anyone can access"
    d
    also access denied, even though I set the link to "anyone can access"
  • It's downloading the file, but something else is wrong
    d
    It's downloading the file, but something else is wrong
  • Now I'm looking into adding "superres_strength" to the synthesis notebook that does work, but it updates its parameters from a list of config defaults in tacotron2.py, and there is no superres_strength in there, so I can't add it... I want to make less mechanical sounding audio...
    d
    Now I'm looking into adding "superres_strength" to the synthesis notebook that does work, but it updates its parameters from a list of config defaults in tacotron2.py, and there is no superres_strength in there, so I can't add it... I want to make less mechanical sounding audio...
  • There's a reason the default epoch amount is 200
    g
    There's a reason the default epoch amount is 200
  • Ok
    t
    Ok
  • could there be a notebook version of this https://app.uberduck.ai/speak for the reference audio tab that i can do locally with just a google colab notebook? load up a localized voice model and then load up some reference audio for it to clone along with the transcription? as mentioned, the site submissions are stalled and I'm seeing if there's any other options to get this done.....the reference audio helps guide the voice to say a better version than just some text alone
    j
    could there be a notebook version of this https://app.uberduck.ai/speak for the reference audio tab that i can do locally with just a google colab notebook? load up a localized voice model and then load up some reference audio for it to clone along with the transcription? as mentioned, the site submissions are stalled and I'm seeing if there's any other options to get this done.....the reference audio helps guide the voice to say a better version than just some text alone
  • > “finalized .pt” insinuates tacotron
    u
    > “finalized .pt” insinuates tacotron
  • > “reference audio” insinuates talknet
    u
    > “reference audio” insinuates talknet
  • which one is it?????
    u
    which one is it?????
  • nvm i see the errors in my ways, didn't know i had to retrain it under a different model
    j
    nvm i see the errors in my ways, didn't know i had to retrain it under a different model
  • > `IndexError Traceback (most recent call last) > <ipython-input-3-1c9a8cb14758> in <cell line: 127>() > 126 dirs = [x for x in os.scandir(target_path)] > 127 if len(dirs) == 1: > --> 128 with open(sorted(glob.glob(os.path.join(dirs[0].path, "*.txt")))[0]) as f: > 129 already_multispeaker = (len(f.readline().split("|")) == 3) > 130 > > IndexError: list index out of range`
    p
    >
    Copy code
    IndexError                                Traceback (most recent call last)
> <ipython-input-3-1c9a8cb14758> in <cell line: 127>()
>     126 dirs = [x for x in os.scandir(target_path)]
>     127 if len(dirs) == 1:
> --> 128   with open(sorted(glob.glob(os.path.join(dirs[0].path, "*.txt")))[0]) as f:
>     129     already_multispeaker = (len(f.readline().split("|")) == 3)
>     130 
> 
> IndexError: list index out of range
  • Someone help me out.
    p
    Someone help me out.
  • false alarm
    p
    false alarm
  • > `/content/project > > --------------------------------------------------------------------------- > > TypeError Traceback (most recent call last) > > <ipython-input-8-7d240b4478d1> in <cell line: 254>() > > 253 > > 254 with open("tacotron2_config.json", "w") as f: > > --> 255 json.dump(config, f) > > 256 > > 257 get_ipython().system('python -m uberduck_ml_dev.exec.train_tacotron2 --config "tacotron2_config.json"') > > > > 4 frames > > /usr/lib/python3.9/json/encoder.py in default(self, o) > > 177 > > 178 """ > > --> 179 raise TypeError(f'Object of type {o.__class__.__name__} ' > > 180 f'is not JSON serializable') > > 181 > > > > TypeError: Object of type complex is not JSON serializable`
    p
    >
    Copy code
    /content/project
> > ---------------------------------------------------------------------------
> > TypeError                                 Traceback (most recent call last)
> > <ipython-input-8-7d240b4478d1> in <cell line: 254>()
> >     253 
> >     254 with open("tacotron2_config.json", "w") as f:
> > --> 255   json.dump(config, f)
> >     256 
> >     257 get_ipython().system('python -m uberduck_ml_dev.exec.train_tacotron2 --config "tacotron2_config.json"')
> > 
> > 4 frames
> > /usr/lib/python3.9/json/encoder.py in default(self, o)
> >     177 
> >     178         """
> > --> 179         raise TypeError(f'Object of type {o.__class__.__name__} '
> >     180                         f'is not JSON serializable')
> >     181 
> > 
> > TypeError: Object of type complex is not JSON serializable
  • when i test alice voice on FakeYou synthesizer notebook
    m
1...139140141...158Latest