• stebo
    link
    fedilink
    711 months ago

    That last thing will still be difficult with AI. You can generate one image that looks convincing, but generating multiple images that are consistent? I doubt it.

      • I feel like you could do this right now by hand (if you have experience with 3d modelling) once you’ve generated an image. 3d modelling often includes creating a model from references, be they drawn or photographs.

        Plus, I just remembered that creating 3d models of everyday objects/people via photos from multiple angles has been a thing for a long time. You can make a setup that uses just your phone and some software to make 3d printable models of real objects. No reason preventing someone from using a series of AI generated images instead of photos they took, so long as you can generate a consistent enough series to get a base model you can do some touch-up by hand to fix anything that the software might’ve messed up. I remember a famous lady in the 3d printing space who I think used this sort of process to make a complete 3d model of her (naked) body, and then sold copies of it on her Patreon or something.