https://uberduck.ai/ logo
Join Discord
Powered by
# machine-learning
  • u

    {K EY1} (Kei)

    10/17/2022, 2:24 PM
    I wish it had something like that too, or an option to train the voice you want but the person said training will remain private and that they're done with tortoisetts
  • u

    {K EY1} (Kei)

    10/17/2022, 8:15 PM
    so i trained a stable diffusion text inversion model for a character called Namine Ritsu
  • u

    {K EY1} (Kei)

    10/17/2022, 8:15 PM
    it didn't work very well...
  • u

    {K EY1} (Kei)

    10/17/2022, 8:16 PM
    (spoiler for bright colors) this is
    <namine-ritsu> playing the piano
  • u

    {K EY1} (Kei)

    10/17/2022, 8:17 PM
    if you want to check it out yourself: https://huggingface.co/sd-concepts-library/namine-ritsu
  • u

    {K EY1} (Kei)

    10/17/2022, 9:26 PM
    ok i trained a model to replicate my art style and it... sorta did it???
  • u

    {K EY1} (Kei)

    10/17/2022, 9:30 PM
    yeah it... sorta looks similar to my art?
  • u

    {K EY1} (Kei)

    10/17/2022, 9:31 PM
    ok it can't draw an apple in my style
  • u

    (Dawn) Will Draw Fictional Women

    10/17/2022, 9:32 PM
    train it on my art that i dont have a lot of other than voice images
  • u

    {K EY1} (Kei)

    10/17/2022, 9:34 PM
    you can do it yourself with this notebook: https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/sd_textual_inversion_training.ipynb
  • u

    (Dawn) Will Draw Fictional Women

    10/17/2022, 9:41 PM
    ill think about it
  • u

    (Dawn) Will Draw Fictional Women

    10/17/2022, 9:41 PM
  • j

    Justin

    10/18/2022, 3:22 AM

    https://youtu.be/fORng0zjXQQ▾

  • m

    mega b

    10/18/2022, 3:26 AM
    dope
  • t

    tanooki426

    10/18/2022, 10:59 AM
    Think you could teach me how to make it draw stuff in art styles from people who previously made art for me?
  • h

    hecko

    10/18/2022, 1:01 PM
    aaa why is it so hard to do outpainting
  • h

    hecko

    10/18/2022, 1:06 PM
    ,,actually it might be a better idea to do the background in ms paint or something and then ask stable diffusion for the head bit
  • h

    hecko

    10/18/2022, 1:17 PM
    btw i'm now running stable diffusion locally and it's surprisingly not that much slower than on t4
  • h

    hecko

    10/18/2022, 1:19 PM
    ooh these new obamnas look quite tempting too
  • h

    hecko

    10/18/2022, 1:40 PM
    i got it to uncrop correctly, then ran it through once more to smooth things out and pfff
  • h

    hecko

    10/18/2022, 1:42 PM
    the uncrop btw
  • j

    Justin

    10/18/2022, 1:43 PM
    What kind of upscaler did you use on it?
  • h

    hecko

    10/18/2022, 1:43 PM
    none
  • h

    hecko

    10/18/2022, 1:43 PM
    just img2img
  • j

    Justin

    10/18/2022, 1:43 PM
    ah iSee
  • h

    hecko

    10/18/2022, 1:43 PM
    because i don't like the contrast and wanted to see if i could fix it
  • j

    Justin

    10/18/2022, 1:43 PM
    mm
  • h

    hecko

    10/18/2022, 1:44 PM
    this is somewhat better but the beak got bleached
  • j

    Justin

    10/18/2022, 1:45 PM
    ye but Obama looks better than before
  • h

    hecko

    10/18/2022, 1:46 PM
    think this might be it
1...994995996...1068Latest