https://uberduck.ai/ logo
Powered by
# machine-learning
  • a
    hold up, that exists?
  • a
    now this is where the fun starts with spongebob voice troll /j
  • s
    Well, I already have some TalkNet voices
  • s
    Hopefully Voicemod is compatible with them
  • u
    i guarantee you its not going to be
  • a
    ^
  • g
    Voicemod is like voice effects and stuff, talknet is like realistic text to speech voices
  • a
    welll true
  • a
    but talknet is not like just realistic
  • a
    you load audio file and it will be realistic
  • a
    or use mic
  • k
    anyone had n brio's roger craig smith voice?
  • u
    testing my Mr. Turner model
  • s
    Does anyone have a pipeline colab with a Spanish base model?
  • u
    just retained him with 20 more samples
  • t
    Uh oh
  • t
    Getting the message "NaN or Inf found in input tensor" with very slowed down progress on training
  • t
    What ought I do
  • h
    uhhhhhhhh
  • h
    one of your neurons either divided by zero or nyoomed up to infinity
  • h
    or one of your datas is scuffed
  • h
    but also that is a Very High Loss
  • h
    you are using a base model right?
  • t
    AITCH
  • h
    hm
  • h
    there must be something seriously wrong with your transcript
  • h
    please confirm that you've disabled arpabet
  • h
    and also send a snippet of the transcription.txt file
  • t
    It took the epoch twice as long to train as usual
  • t
    Copy code
    "Other": {
    "symbol_set": "nvidia_taco2",
    "text_cleaners": ["transliteration_cleaners"],
    "p_arpabet": 0.0,
    "sample_inference_text": ""
1...939940941...1068Latest