https://uberduck.ai/ logo
Join Discord
Powered by
# machine-learning
  • w
    And I do all of my synthesis locally so that would take a lot of it
  • w
    This was supposed to be twice a WEEK not month
  • a
    Copy code
    js
Window size: 320
Model: Vocal_HP_4BAND_3090_arch-124m
Parameter: 4band_44100
Aggressiveness: 0.3
High end process: mirroring2
TTA: True
Deep Extraction: False

loading model... Model loading failed, trying again...
loading model... Model loading failed, trying again...
loading model... Model loading failed, trying again...
loading model... Model loading failed, trying again...
An error has occurred: [Errno 2] No such file or directory: 'models/v5_new/Vocal_HP_4BAND_3090_arch-124m.pth'
---------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
/usr/lib/python3.8/shutil.py in move(src, dst, copy_function)
    790     try:
--> 791         os.rename(src, real_dst)
    792     except OSError:

FileNotFoundError: [Errno 2] No such file or directory: 'separated/1_Vocal_HP_4BAND_3090_arch-124m_Instruments.wav' -> '1/1_Vocal_HP_4BAND_3090_arch-124m_Instruments.wav'

During handling of the above exception, another exception occurred:

FileNotFoundError                         Traceback (most recent call last)
4 frames
/usr/lib/python3.8/shutil.py in copyfile(src, dst, follow_symlinks)
    262         os.symlink(os.readlink(src), dst)
    263     else:
--> 264         with open(src, 'rb') as fsrc, open(dst, 'wb') as fdst:
    265             # macOS
    266             if _HAS_FCOPYFILE:

FileNotFoundError: [Errno 2] No such file or directory: 'separated/1_Vocal_HP_4BAND_3090_arch-124m_Instruments.wav'
  • a
    how can i fix it?
  • h
    uhhh good question, it seems the file host they used for the model broke i propose https://mvsep.com/
  • a
    But do I have to wait in the queue?
  • a
    I've used this site before and it takes forever to wait
  • h
    odd, every time i used it it was pretty quick
  • h
    you can also run 
    Ultimate Vocal Remover
    on your own computer but it takes up several gigabytes
  • a
    Is it from Colab?
  • h
    no
  • h
    colab only hosts the thing
  • h
    but it's the same model as on colab
  • t
    I finally have the voice data I need for Elmer. Of course I transcribe the first line as: "Shh! Be very quiet, I'm hunting rabbits." Right?
  • a
    oh ok
  • a
    thanks
  • h
    yeah
  • t
    At least he's FINALLY going to say Shhh! without the messed up transcription that has "Shh!" removed from Jacob's version. Jacob if you're reading this, no offense dude, but I think that model needs some work and I'm here to help.
  • j
  • m
    Anyone know a TTS library where you can dictate the rhythm of the output using alphanumeric characters, instead of a reference track?
  • m
    Something that could be called from a CLI like this:
  • m
    /path/to/hypotheticalTTSbinary --output_text "Hello world" --output_rhythm 101
  • m
    The 1's indicate emphasized syllables, and the 0 indicate non-emphasized syllables
  • m
    I basically need a TTS library whose outputted .wavs have rhythms I can control programmatically by some method other than a reference track
  • m
    A package that implements Speech Synthesis Markup Language (SSML) might work: https://www.w3.org/TR/speech-synthesis11/
  • h
    you may be looking for arpabet
  • h
    it's supported by some of the voices on uberduck
  • h
    you can convert with https://app.uberduck.ai/g2p
  • h
    and then adjust the numbers, 
    1
    is stress 
    0
    is no stress 
    2
    is half stress
  • u
    @hecko quacked: hello world. {HH AH0 L OW1} {W ER1 L D}. {HH EH1 L OW0} {W ER1 L D}.
1...102610271028...1068Latest