Show some recent work

Out of context: Reply #8604

  • Started
  • Last post
  • 8,632 Responses
  • microkorg10

    Some outputs from a new generative WIP.
    Code controlled AI generated imagery + Code drawing shapes, scribbles and overlay effects

    microkorg 10Permalink
    UpvoteDownvote
    Flag
    • +++++++PhanLo
    • fuuuuuumonospaced
    • Going to attempt to introduce white borders and within then at random places have print registration marks and cmyk squares.microkorg
    • Yes!maquito
    • Why add fake registration?monospaced
    • Why do you think you have anything to add, monospaced? Shut the fuck upimbecile
    • excuse me?monospaced
    • in regard to design, I probably have more to add than youmonospaced
    • i said shut up, you stupid fuck. no one asked for your input or critique.imbecile
    • Couldn't get registration to work via code but have now got a white frame around the outputs that is equal all around but varies in width slightly per output.microkorg
    • lol, very mature dialog going on here.
      besides that, the images look very cool.
      Code controlled means no prompt or input?      Beeswax
    • Very nice @microkorgmort_
    • Thanks. The typographic and photographic imagery is all prompt based outputs and they are being scaled, rotated, placed randomly by code.microkorg
    • The image outputs are created in StableDiffusion using prompts with variables injected plus influence from.previous typographic work on MidJourneymicrokorg
    • Very coolskinny_puppy
    • NoiceRamanisky2
    • I assume you're possibly generating hundreds or more and selecting what you believe to be the best?kalkal
    • Most of these were one after the other outputs. Maybe skipped a couple of not as good examples. They take about 10 seconds to render each when hitting refresh.microkorg
    • Not got the patience to go through hundreds.microkorg
    • Adding a white border is looking good. Have updated the code so there's less chance of 'star' shapes with the vectors by forcing more obtuse & reflex angles.microkorg
    • Plan to give it far more image and type assets to play with too.microkorg
    • reckon I could have a crack at something like this, not a coder but with chatgpt, or the new claude model, seems even better. it's always sucked at blendingkalkal
    • modes in python thokalkal
    • These are cool but notice how AI is good at creating abstract or deconstructed looking images like these. But try to get it to do clean, minimal layouts andyuekit
    • the results are less impressive in my experience. Not sure if that's an inherent weakness of generative AI or if it's just a product of how existing systems areyuekit
    • trained.yuekit
    • The AI being used in this process is only in the creation of the image and type assets. These are fixed jpegs/gifs.
      It is the code that's doing the composition      microkorg
    Show [[ numHiddenNotes ]] more notesAdd Note
    SaveCancel

View thread