https://uberduck.ai/ logo
Powered by
# machine-learning
  • s
    yeah might have to fix the dataset and redo the training
  • s
    more training, same issue :/
  • a
    You can split some of the samples. 89, 90, and 96, for example, have two sentences. Would making each sentence their own clip work?
  • s
    yeah that would be a better idea
  • s
    damn, hate to lose all this progress
  • a
    If you split 89, for example, you can rename the resulting portions into 89a and 89b and leave the others with their original filenames.
  • s
    if i have to split lines i just add them to the end of my dataset
  • s
  • s
    one more
  • j
    You gotta input a video...
  • u
    It worked
  • u
    Thx
  • j
    open file browser
  • j
    A'ight
  • w
    Lol
  • w
    This is horrible
  • m
    For Larry, the sounds I got from the Sound Resource website, the audio data is 8:55.
  • u
  • p
  • m
    dam daniel
  • m
    thats funny lol
  • m
    will try with dall-e 👍
  • s
    that's unreal
  • u
    Someone generated a image that almost looks like Elon musk
  • m
    Not using DALL-E: 
    walking down an empty corridor. unreal engine
  • p
    if you have capacity on your colab to spare you guys could run this colab (just open, click runtime and run all) https://colab.research.google.com/drive/1P1H-1kc_CFgJE1NOnXywm2fVSoyv2gMW?usp=sharing
  • m
    ^
  • m
  • p
    it's an effort to build a large public image dataset
  • m
    About to hit 200,000 scraped pairs soon!
1...182183184...1068Latest