https://uberduck.ai/ logo
Join Discord
Powered by
# talknet-support
  • w

    WeegeeFan1

    11/20/2022, 2:56 AM
    I've heard there is a simple way to uprgade it
  • w

    WeegeeFan1

    11/20/2022, 2:56 AM
    You simply change a text in the setup stuff for it or something
  • a

    AlexTheAce

    11/20/2022, 2:56 AM
    inside the page itself?
  • w

    WeegeeFan1

    11/20/2022, 2:56 AM
    Did you do that?
  • w

    WeegeeFan1

    11/20/2022, 2:56 AM
    Yes
  • w

    WeegeeFan1

    11/20/2022, 2:56 AM
    Scripts
  • w

    WeegeeFan1

    11/20/2022, 2:57 AM
    That's for talknet at least
  • w

    WeegeeFan1

    11/20/2022, 2:57 AM
    On some very old models
  • w

    WeegeeFan1

    11/20/2022, 2:57 AM
    But the notebook SHOULD be updated already
  • w

    WeegeeFan1

    11/20/2022, 2:57 AM
    at least the ones that the #903688200973873222 channel uses
  • a

    AlexTheAce

    11/20/2022, 2:57 AM
    Yeah, I tried but I got this instead Colab only includes TensorFlow 2.x; %tensorflow_version has no effect.
  • w

    WeegeeFan1

    11/20/2022, 2:58 AM
    I don't really know..
  • a

    AlexTheAce

    11/20/2022, 2:59 AM
    do you know anyone who does?
  • w

    WeegeeFan1

    11/20/2022, 3:00 AM
    Hmm give me a second..
  • a

    AlexTheAce

    11/20/2022, 3:00 AM
    I've been asking around and I got little to no help at all
  • w

    WeegeeFan1

    11/20/2022, 3:00 AM
    CALLING OUT TO @hecko! Do you know anything about tacotron 2 notebooks? We have @AlexTheAce here with a question that is most likely very simple to answer about a tacotron notebook.
  • n

    notcharldeon

    11/20/2022, 3:14 AM
    GUYS what happened to the pitch multiplier option
  • n

    notcharldeon

    11/20/2022, 3:15 AM
    from the manual
  • n

    notcharldeon

    11/20/2022, 3:15 AM
    it's way too low
  • a

    AlexTheAce

    11/20/2022, 4:08 AM
    @hecko Since you're good with tacotron, can you help me with this? /content/project --------------------------------------------------------------------------- ValueError Traceback (most recent call last) in 124 with open("/content/project/transcription.txt") as f: 125 lines = f.read().splitlines() --> 126 speaker_count = max([int(i.split("|")[2]) for i in lines]) + 1 127 wav_count = len(lines) 128 ValueError: max() arg is an empty sequence
  • h

    hecko

    11/20/2022, 12:19 PM
    this is the talknet channel for one
  • h

    hecko

    11/20/2022, 12:20 PM
    for two i think your transcript has issues
  • h

    hecko

    11/20/2022, 12:20 PM
    might even be empty
  • h

    hecko

    11/20/2022, 12:20 PM
    not necessarily the one you gave it but the one it generated
  • m

    moonchild

    11/20/2022, 7:49 PM
    I don't have enough sources where my model is singing to use it for the dataset, so can I use some lines where she is talking and use the same songs after changing the pitch to increase the dataset?? would it make any difference??
  • h

    hecko

    11/21/2022, 12:11 AM
    talking: if the tone of voice is close enough pitch change: go for it
  • w

    WeegeeFan1

    11/22/2022, 2:23 AM
    https://drive.google.com/file/d/1c_MvSh_yDoVbwENtHjPDldSG6GODDYjY/view?usp=sharing
  • w

    WeegeeFan1

    11/22/2022, 2:24 AM
    So I have a model and I haven't been getting amazing results. I cannot tell if it's because of my dataset or if it's because of me not training it long enough. Please someone look and tell me.
  • w

    WeegeeFan1

    11/22/2022, 2:25 AM
    The way I've organized the zip is everything without "cab[#]" before it is audio of interviews of him. Everything with "cab" before the name is singing data, which is the majority.
  • r

    Radak

    11/22/2022, 6:30 PM
    Can a model be trained off of three different languages?
1...434445...74Latest